Version and Change Control
Every change
Each stall keeps its own version and its own log. Every edit to the words or numbers of a document is recorded before it is released: the sections it touches, what it was before and after, why, and who approved it.
Until 1.10 the two stalls shared one version and one log. In 1.11 that log was split: each row went to the stall whose documents it touched, and to both when it touched both. So the shared history appears in both logs below.
Stall 01
The Hawker Library
Version 1.11 · 2 November 2026 · tag library-v1.11
A · The Record · C · Hawkers and Social Protection · D · Follow the Plate · E · Succession Economics · F · Method, Levers and Data · G · The Technical Paper · S · Sources · Your Table. 15 versions and 350 changes recorded.
↓ Download 0L · the Library Change Log (PDF, 505 KB, 200 pages) The Library
The PDF also holds the development history before 1.0 and the SHA-256 of every released PDF.
1.11 2 November 2026 4 changes
Its own version. The Hawker Library now has its own version, history, tag and Version and Change Control Log (0L, the Library Change Log), starting at 1.11. Until 1.10 it shared one version and one log with Saving our Hawker Heritage. On the website, every Library page now lives under /library/. No proposal, figure or claim changes.
Approved by Uncle James (2 Nov 2026) · tag library-v1.11
Before. One version for all twelve documents, 1.10, tagged v1.10, and one Version and Change Control Log, 0 (Hawker_Heritage_0_Change_Log.pdf), for both stalls
After. The Hawker Library has its own version, 1.11, tagged library-v1.11, and its own log, 0L, the Library Change Log (Hawker_Heritage_0L_Library_Change_Log.pdf). Each of its documents prints that version on its cover, in its footer, in its metadata and on its Document Control page, which now names the stall and its log. The shared history, 1.0 to 1.10, is in both logs: each row of it sits in the log of the stall whose documents it touched, and in both when it touched both. The website's /changes/ page shows both logs
Why. Uncle James's decision of 2 Nov 2026: each stall keeps its own version and change log, so a change to one stall no longer moves the other's version. A restructure by the initiator alone, so a minor version
Before. "the twelve documents", "all twelve documents", "one of the twelve documents" (since 1.10, for every document of Hawker Heritage with its one change log)
After. "the thirteen documents", "all thirteen documents", "one of the thirteen documents": the eleven documents and the two change logs, 0L and 0P
Why. With a change log for each stall there are thirteen documents, not twelve; no other word changes
Before. "which does all of it and composes the twelve documents"; "Every change from version 1.0 is recorded in the Version and Change Control Log."
After. "composes the thirteen documents"; "Every change from version 1.0 is recorded in the Library Change Log, the Hawker Library's own Version and Change Control Log (0L)."
Why. The same count, and G's own log, in the canonical text of G.9; no other word changes
Before. "walk through a day at one (hawker.siliconwok.com/day)", linked to /day/
After. "walk through a day at one (hawker.siliconwok.com/library/day)", linked to /library/day/; the old address redirects
Why. Since 1.11 every Library page on the website lives under /library/, so that www.siliconwok.com can show the same pages (Uncle James, 2 Nov 2026)
1.10 2 November 2026 8 changes · shared version
Two stalls. The documents now sit in two series, the Hawker Library (the evidence) and Saving our Hawker Heritage (the Plan), each by Uncle James, a Silicon Wok programme. The Library adds a graded estimate of what hawkers earn above the median, and the Plan uses it in three places. No proposal, ask or model figure changes.
Approved by Uncle James (2 Nov 2026) · tag v1.10
Before. The cover's top line "THE PLAN · INITIATED BY UNCLE JAMES" (and the same for the Executive Summary, "COMPANION VOLUME A" to G and "THE SOURCES"); the companion volumes' series line "Saving Singapore's Hawker Heritage · a library of standalone documents"; the Executive Summary's and the Plan's cover title "Saving Singapore's Hawker Heritage"; Your Table's masthead line "SAVING SINGAPORE'S HAWKER HERITAGE · YOUR TABLE"; every page's running footer "SAVING SINGAPORE'S HAWKER HERITAGE · …"; every document's title "Saving Singapore's Hawker Heritage — …" in Document Control and the PDF metadata
After. Two series. The Executive Summary, the Plan and the Budget Book carry "SAVING OUR HAWKER HERITAGE · … · BY UNCLE JAMES · A SILICON WOK PROGRAMME", the cover title "Saving our Hawker Heritage" (the Executive Summary and the Plan) and the title "Saving our Hawker Heritage — …". Volumes A and C to G, Sources and Your Table carry "THE HAWKER LIBRARY · … · BY UNCLE JAMES · A SILICON WOK PROGRAMME" and the title "The Hawker Library — …". The companion volumes' series line is the series' own line: "The Hawker Library · The open reference on Singapore's hawker trade: its history, its prices, its incomes and its sources." (B: "Saving our Hawker Heritage · A citizen's plan, built on the Library. …"). Each running footer names the series: "THE HAWKER LIBRARY · FOLLOW THE PLATE", "SAVING OUR HAWKER HERITAGE · THE PLAN"
Why. Uncle James's decision of 2 Nov 2026: split the work into a Hawker Library, the evidence, which never argues, and a Saving our Hawker Heritage Plan, built on it. Only the connective cover and series text changes; the canonical text is not retitled
Before. G.3.2 gave the model's net income as deciles, quartiles and the mean, to the top tenth ($5,140 a month, all stalls), with no word on the top end
After. A short box, "Why the model has no top tail": the model lets plates a day vary by only about 15% either way, so it has no top tail; its middle agrees with the outside evidence; on graded evidence one stall's household may take home $8,000–12,000 a month at the top 5% and $15,000–30,000 at the top 1% (2026 prices), a plausible range, not a measurement; a pointer to the Library's page What hawkers earn. The model's numbers do not change
Why. The income findings of 2 Nov 2026 ("Hawker incomes above the median"): the model's middle is right, but it has no tail. Graded ranges, not a new model figure (Uncle James, 2 Nov 2026)
Before. "… and 7¢ to 21¢ reaches the hawker's family. On a 13-hour day, …"
After. "… reaches the hawker's family. These are typical stalls of each kind. The best-selling 1% of stalls take home several times what a typical stall does (What hawkers earn, in the Library). On a 13-hour day, …"
Why. The hawker's share is for a typical stall, not every stall; the income findings show a long top tail
Before. "Pay the goodwill." … "It is never priced on the location, which is the public's, and NEA may cap it. The veteran finally has a five-figure reason …"
After. Between those sentences: recipe sales show how wide the range runs, from $20,000 for a well-known Hokkien mee stall's stall and recipes (2020) to $500,000 to $1 million asked or offered for a few famous stalls' (2022–23); prices paid or asked, not measured incomes (grade C); they support a capped payment held in escrow, and give NEA a basis for where to set the cap
Why. Recipe prices give the cap on the goodwill payment a basis; no change to the ask
Before. 457 references; the Technical Paper's group "The Technical Paper's Series"
After. 474 references: 17 new ones behind the income estimate (Mothership, IPS, a Member's speech, Al Jazeera, Yahoo, Malay Mail, Vulcan Post, Tabla, SingSaver, KitchenUnionSG and a Reddit thread supplied as screenshots), and G.3 added to six existing ones (MSE Apr 2022 and Sep 2024, LKYSPP 2024, MTI 2015, TODAY 2024, Home & Decor 2026); each record's note gives its grade, A to D. The group is "First Cited in the Technical Paper". On the website, a new Library page, What hawkers earn (/library/earnings/, from data/curated/earnings.json), and a Glossary entry, "Evidence grades"
Why. Every number on the new page and in the new lines traces to a source the Library prints, with its grade kept visible
Before. The old umbrella title, "Saving Singapore's Hawker Heritage", on the change log's cover and title ("Saving Singapore's Hawker Heritage — Version and Change Control Log") and in its opening note; Document Control's independence line "Hawker Heritage is an independent initiative stewarded by Silicon Wok". Outside the PDFs: the website's home heading and page title, the dataset name on the download page and the film's label, README.md, CITATION.cff, the simulator's eyebrow and note, the model's docstring and the canonical sources' titles
After. "Hawker Heritage, a Silicon Wok programme" (or "Hawker Heritage" where space is short): the change log is "Hawker Heritage — Version and Change Control Log", "one of the twelve documents"; the independence line reads "Hawker Heritage is an independent Silicon Wok programme", as on the website. The stall names stay: Stall 01 the Hawker Library, Stall 02 Saving our Hawker Heritage (the Plan). The website's home section is "Three stalls"
Why. The old umbrella title is retired; the collection is Hawker Heritage, a Silicon Wok programme, and the two disclaimers now read alike (Uncle James, 2 Nov 2026)
Before. Lower-case "the library" for all twelve documents: "Document 1 of the library", "Document 2 of the library", "document 1 of the library", "Volume G of the library"; "published as a small library with three doors" (ES.1); Document Control's row "The library" and its note "published beside the library"; "the references for the whole library", "for the library", "for every document in the library", "Sources, the library's list of references" (the sources notes of the Plan and of A to G, F.5); "where every lever in the library is printed" (the levers); "The library's text" (F.4); in S, "every reference the library cites", "in the library's order", "where in the library it is used", "Three marks run through the library", "found a link in the library", "any later document of the library"; the descriptions of G ("behind the library") and S ("Every reference the library cites")
After. "Document 1 of Hawker Heritage" and the same for 2 and G; "published as twelve documents with three doors"; the row "Hawker Heritage", "one of the twelve documents"; "for all twelve documents", "Sources, their list of references" or "the list of references for all twelve documents"; "every lever in the twelve documents"; "The documents' text"; in S, "every reference the twelve documents cite", "in Hawker Heritage's order", "where in the twelve documents it is used", "run through the twelve documents", "a link in Hawker Heritage", "any later document of Hawker Heritage"; "behind the twelve documents", "Every reference the twelve documents cite"
Why. Since 1.10 "the Library" is Stall 01; "the library" for all twelve documents now reads as that stall. Where it really means the Hawker Library it stays (Uncle James, 2 Nov 2026)
Before. "Anyone can rebuild every number in this library from the repository"; "which does all of it and composes the library"; "This volume belongs to version 1.10 of the library"
After. "every number in Hawker Heritage"; "composes the twelve documents"; "version 1.10 of Hawker Heritage"
Why. The same swap in the canonical text of G.9; no other word changes
1.9.3 1 November 2026 3 changes · shared version
The time-bound body was the UNESCO nomination committee of 2018–20, not FMAS. FMAS is the trade's standing federation and carries on. The library now says exactly what MCCY was asked and answered on 7 Oct 2026, and speaks of NEA and its Hawker Centres Group as the partners whose work it builds on. No proposal, figure or claim changes.
Approved by Uncle James · tag v1.9.3 · release commit 9e7498e
Before. Two owners, one table (§3.8; A.5, B.14): the 14-member committee that won the inscription "lapsed" with it. The ask for a permanent committee (§7.7, B.7, ES.6, the Plan's short form, A.7): "the 2018–20 nomination committee of NHB, NEA and FMAS", "brought back" or "asked back as a permanent body". The A.5 opener: "the committee that won the inscription and then lapsed"
After. "co-chaired the 14-member UNESCO nomination committee that won the inscription. That committee was set up for the nomination alone, and the public record shows no meeting after the inscription. FMAS itself carries on, as it has since 1989." The ask names "the 2018–20 UNESCO nomination committee", co-chaired by NEA, NHB and FMAS, "brought back as a permanent body", and says FMAS itself is a standing federation. The A.5 opener: "the UNESCO nomination committee that won the inscription, set up for that task alone"
Why. Make plain that the time-bound body was the UNESCO nomination committee, not FMAS, which is a standing federation (Uncle James, 1 Nov 2026).
Before. §3.8 (A.5, B.14): asked "whether a formal mechanism should be set up", the Government "described the two agencies' work side by side" and, the line said, named no mechanism. §1.1 (A.4): "no count of practitioners or of dishes at risk (MCCY, 7 Oct 2026)". The website's card for the permanent committee: MCCY named no mechanism of a joint kind
After. "Asked in October 2026 whether a formal mechanism should identify the trades at risk, MCCY pointed to the schemes already running and left the question open (MCCY, 7 Oct 2026)." §1.1 (A.4): "no list of trades or practices at risk (MCCY, 7 Oct 2026)". The card: the same wording as §3.8
Why. Say exactly what MCCY was asked and answered on 7 Oct 2026 (Uncle James, 1 Nov 2026)
Before. §1.1 (A.4): "A scheme that reports enrolment is reporting its budget"; "Both are true; only one is reported"; "He may charge what he likes, as long as it is $3.50"; "The blank is on UNESCO's record too: … named the threats and gave no count". A.2.1: "a token stipend". A.4, D.14: "for what NEA could not do". B.7: "The number Parliament was told does not exist". ES.1: "The official reassurances … Four of them leave out the part that matters". Ideas credited to a party: the ballot "the Progress Singapore Party proposed in Parliament in 2024" (B.7), "credited to the PSP's 2024 proposal" (A.7), "Two Progress Singapore Party MPs" (ES.5, §1.3), and party labels on the Members credited in A.3, D.14, D.15 and F.9
After. "A count of outcomes would show what works"; "Both are true, and both are NEA's own findings"; "He sets his own prices, but the $3.50 value meal needs approval to change"; "named the threats but could not yet count them; the next one could"; "a small stipend"; "for something a price tender cannot do"; "The count Singapore does not yet keep"; "The official figures … Four of them need a second number beside them". "the ballot proposed in Parliament in 2024"; "credited to a proposal made in Parliament in 2024"; "Two Members, Leong Mun Wai and Hazel Poa"; the Members are named without their parties. The sources, Hansard and the parties' own published speeches among them, are unchanged
Why. A constructive tone towards NEA and its Hawker Centres Group, whose work the plan builds on; no change to any fact, figure or ask (Uncle James, 1 Nov 2026)
1.9.2 1 November 2026 2 changes · shared version
The pen name on every cover. 1.8 changed the byline, Document Control and the metadata, but the top line of every cover still gave my full name; it now reads Uncle James. No proposal, figure or claim changes.
Approved by Uncle James · tag v1.9.2 · release commit d28eda7
Before. The cover's top line: "THE EXECUTIVE SUMMARY · INITIATED BY", "THE PLAN · INITIATED BY", "COMPANION VOLUME A · INITIATED BY" (to G) and "THE SOURCES · INITIATED BY", each followed by the initiator's full name
After. "… · INITIATED BY UNCLE JAMES" on every cover
Why. the author's request: his pen name on every PDF. 1.8 changed the byline, Document Control and the PDF metadata, but not this line
Before. Two past rows still named him: the 1.3 row on §3.5's retention target ("for" his first name "to confirm") and the 1.7 row on "the plan" (the Plan's cover line, with his full name)
After. "TARGET CHANGE FOR UNCLE JAMES TO CONFIRM" and "THE PLAN · INITIATED BY UNCLE JAMES". No other past row changes, and the released PDFs' fingerprints are untouched
Why. the author's request: his pen name, in past rows too
1.9.1 1 November 2026 1 change · shared version
The site's address and the contact in the papers are now Silicon Wok's. No proposal, figure or claim changes.
Approved by Uncle James · tag v1.9.1 · release commit 9d75eca
Before. hawker.silicontribe.ai
After. hawker.siliconwok.com
Why. the author's decision: Silicon Wok is the publisher and contact
1.9 1 November 2026 1 change · shared version
Hawker Heritage is now stewarded by Silicon Wok, the home of my writing as Uncle James. No proposal, figure or claim changes.
Approved by Uncle James · tag v1.9 · release commit 3c99397
Before. "Hawker Heritage is an independent initiative stewarded by Silicon Tribe."
After. "Hawker Heritage is an independent initiative stewarded by Silicon Wok."
Why. the author's decision: Silicon Wok is the initiative's home
1.8 31 October 2026 6 changes · shared version
A pen name and a shorter biography. From now on I write as Uncle James, and the library, the site and this log carry that name. The note from the author no longer names the places I have worked. No proposal, figure or claim changes.
Approved by Uncle James · tag v1.8 · release commit d6fcec9
Before. Initiated by, and the PDF author: the initiator's full name
After. Uncle James (versions.json people.initiator), in every document and in the change log
Why. the author's request: pen name and an anonymised biography
Before. "For diners · from" and the signature in the author's full name
After. "For diners · from Uncle James"; signed "Uncle James · October 2026"
Why. the author's request: pen name and an anonymised biography
Before. A plan initiated by the author's full name, October 2026
After. A plan initiated by Uncle James, October 2026; the same in the design system's Masthead preview
Why. the author's request: pen name and an anonymised biography
Before. About named him in full, with a bio naming his universities, the government agency, the venture firm and its co-founder, the companies he founded and helped build, the acquirer, his employer and a board seat; citations read "Surname, I."
After. "…a plan initiated by Uncle James…"; the bio reads: "Uncle James is based in Singapore. He trained as an engineer and studied public policy. He began his career in the Singapore Government, then spent years as an early-stage venture investor and a founder, most recently of an electric-motorcycle company whose technology and team were later acquired. He now works in the motor industry, and writes as Uncle James at siliconwok.com." The next paragraph is kept. Citations name the pen name as written: "Uncle James (2026)." The same in the build brief (website/BRIEF.md §7.8, §12) and the mock-up. The film's end card reads "Initiated by Uncle James" in its source; the rendered film is not rebuilt in this release
Why. the author's request: pen name and an anonymised biography
Before. Copyright, suggested credit, citation (surname and initial) and "initiated by" in the author's full name
After. "Copyright (c) 2026 Uncle James", a pen name; the licence terms are unchanged. CITATION.cff names the author by alias, "Uncle James"; README cites "Uncle James (2026)" at version 1.8
Why. the author's request: pen name and an anonymised biography
Before. approver and decided_by in the author's full name in every entry, and his first name in past rows' reasons; the 1.1 row on voice named the company whose updates were studied
After. Uncle James in every entry, past ones included; the 1.1 row reads "his blog, essays and company updates". No other past row changes, and the released PDFs' fingerprints are untouched
Why. the author's request: pen name and an anonymised biography
1.7 30 October 2026 26 changes · shared version
Plain words. Three readers found 841 places in the library where a diner, an MP's aide or a journalist would have to read twice, and 285 where they could not tell what was meant. I settled the nine numbers that did not add up first, then applied eight fixes by rule – one dash, one name for the plan, a name beside every id, a gloss at the first use of every term, the metaphors and the code out, long cells cut to lines, one sense per word – and rewrote the 285 by hand. No proposal, figure or claim changes.
Approved by Uncle James · tag v1.7 · release commit a99b0dc
Before. the three together add about $407 a month to the hawker's profit, or $4,791 a year in income and savings
After. about $407 a month to the hawker's profit before tax, or $4,878 a year; after tax, $4,791 a year in income and savings
Why. Plain-English audit: $407 × 12 is not $4,791. Both are the model's and measure different things: profit before tax ($406.52 a month, $4,878 a year) and the household's income and savings after $87 of tax ($4,790.95). Uncle James, 30 Oct 2026, 06:43: the eight rule-based fixes approved
Before. some 230 termination notices between 2012 and 2024 ..., and in 2024 alone more than 230 warnings and 30 terminations
After. termination notices to about 230 stallholders between 2012 and 2024, each after earlier warnings; in 2024 alone letters of warning to more than 230 and termination notices to more than 30
Why. The same figure twice read as an error. Both are MSE's (22 Sep 2025) and count different things; the text now says which
Before. a fee to about 1.25 participants for every cooked-food stall, the 150 of the 120-stall centre
After. a $150 fee to about 1.25 participants for every cooked-food stall (150 participants at the 120-stall centre the ledgers were first drawn for)
Why. "The 150 of the 120-stall centre" ran the fee and the head count together: $150 a participant, 1.25 a stall, 150 people at 120 stalls (the ledger's rates)
Before. the framework 2.1 terms; the 2.1 terms do not
After. the capped terms (the 30% cap and $600 floor, before the protected income); the capped terms do not
Why. "The 2.1 terms" named a version of the 1.0 framework. They are the capped terms, which the rest of E.1 already printed
Before. about 936 plates a week of noodles and rice, or 720 orders at $5.85, 50, 100 and 250 regulars place 10.4%, 20.8% and 52.1% of its orders
After. Take the typical tendered stall. It sells about 936 plates of noodles and rice a week, which is 720 orders of $5.85 each. If it has 50, 100 or 250 regulars, they account for 10.4%, 20.8% and 52.1% of its orders
Why. The run of numbers could not be parsed; printed as the audit's rewrite
Before. em dashes (—) in running text, about 900 of them
After. spaced en dashes ( – ), with an unspaced one for an empty cell; check_layers.py now fails on any em dash in reader text outside an official title that has one
Why. The house style is spaced en dashes. Plain-English audit, fix 1
Before. the framework, this framework; THE PLAN · A POLICY FRAMEWORK on the Plan's cover
After. the plan, this plan; THE PLAN · INITIATED BY UNCLE JAMES
Why. One name for one thing. "Framework" stays only where it names the 1.0 document and its versions. Fix 2
Before. a cross-reference printed its id alone ("see D.4.1")
After. the first time an id appears in a section of prose it carries its section's name, in italic ("D.4.1, How much rent can a stall bear?"); the bare id after. Not in tables, cards, contents, pointers or runs of ids
Why. Readers need the thing's name, not only its id. Fix 3, at build (src/layers/plain.py)
Before. abbreviations and policy terms (HDP, AMR, median, tranche, escrow, means-tested ...) used without a gloss
After. a plain gloss in brackets at the first use of each in each document, from the glossary (113 terms, 68 of them new); the site's pages carry the same
Why. Fix 4, at build (src/layers/plain.py; data/curated/glossary.json gains gloss and abbrev). The Executive Summary stays at eight pages
Before. addon_programme, centres_vetoed, four_centres, sample_stall(), plate_model.pnl(), centre_ledger(), run_handover(), PARAMS keys in F.3; ASSUMPTION in capitals in B.9
After. plain words (each centre's sizing; the model's profit and loss; the funding model's ledger ...); F.3 names each parameter; "assumption" in lower case in B.9 and the funding notes
Why. Code out of prose; the identifiers stay in G.1's register. Fix 8
Before. rails that stop one station short; the engine on the rail; plumbing and pressure; the incubation runway extended to touchdown; a bench of locum cooks; a rent cliff and a glide to band; trust deposits
After. plain words: each scheme stops short of the cook's income; the money the HSS lacks; a pool of relief cooks; the rent phased in over three years; gifts first. §2.3 is now "The Hawkers Succession Scheme, Extended – Pay the Veteran for the Goodwill". "Lighthouse" stays, explained at its first use as the first two pilot centres
Why. Plain-English audit, fix 5: a metaphor that needs the policy context to land. Uncle James, 30 Oct 2026, 06:43: the eight rule-based fixes approved, with the twenty rewrites of the audit as written
Before. table cells of 100 to 500 words, one run-on sentence each; D.7.4's note of about 480 words in one block
After. one fact or one action a line in every long cell; §6.1's Year 0 steps numbered in the order they happen; D.7.4's note a short list under "How to read the national table"; B.9's assumptions table prints one sentence a line
Why. Plain-English audit, fix 6: dense table cells become short lines
Before. "floor" for an income guarantee, a minimum payment, a wage and a stall's site; "tier" for the two tiers, for Bronze to Platinum and for centre Tiers A–C; Band and Tier for the same rent letters; goodwill as kindness; a goodwill sum "retired"
After. the floor is the Protection Floor only (an income top-up, a $600 minimum, the Progressive Wage, site costs); Bronze to Platinum are grades; centres are Class A–C and the rents Bands A–C; goodwill is money only; a goodwill sum is paid off; "a stall with long queues" for a strong business
Why. Plain-English audit, fix 7: one sense per word
Before. 106 sentences the audit marked High (98 in the Plan's sections, 8 in the ledgers' row labels), among them the one-sentence ballot record of §1.3.2, the gate criteria of §6.5, the safeguards of §4.6 and the "stock and flow" of §7.8
After. each rewritten, the ballot record in three blocks (what was proposed, what the Government said, where I stand), the safeguards and the gate as numbered lists, every proposal card and the ask table in the same plain words; the curated proposal data quotes the new text
Why. Plain-English audit: the High rows, in reader order. The Plan ends at 46 pages, as with the glosses alone; Uncle James accepted 45 or 46 with the full glosses (30 Oct 2026)
Before. the volume openers' 15 High rows: "Government's rails come first", "the record is a decade of right rails without engines", "it survives a 2020", "the reserve must cover the bad tail"
After. plain bottom lines; the Budget Book's rise of $4,693,468 in four numbered reasons; CCCS, FSI, PMET and O&M spelled out
Why. Plain-English audit: the High rows, in reader order
Before. 30 High rows, among them "a contributing counterparty", "flows on", "little to annuitise", "the incidence rule", a 330-word note and a 150-word list of open questions
After. someone who pays a share of his CPF; spills over into retirement savings; C.6 "Enough to Retire On? Little to Turn into Monthly Income"; who ends up with the money; the note in four short paragraphs; the questions numbered; "bloc" becomes "group" throughout
Why. Plain-English audit: the High rows, in reader order
Before. 108 High rows in the model's prose and the levers, among them "custom", "recapture", "a lossy transfer", "a regression ... each driver scaled to one standard deviation", "a revolving reserve ... the P95 payout of the open book", a 400-word paragraph in D.5.4 and the funding model's "draw, close and bond ladder" undefined
After. customers; the landlord takes it back at renewal; the price rise loses $3,140 on the way, to customers who stop coming and tax; the factors and a typical-sized step; a reserve that is topped up, with P95 explained; D.5.4 in short paragraphs; B.9 defines its four terms before it uses them; the 5,000 futures as a list
Why. Plain-English audit: the High rows, in reader order. Every number is the model's, unchanged
Before. nine High rows, among them "halfway to the stall's assessed market rent for three more", "drawn again with the same seed, as deciles, quartiles and means", "undiscounted"
After. plain words for each, with the seed, the contingency and the time value of money explained
Why. Plain-English audit: the High rows, in reader order
Before. ASSUMPTION, ASSUMED, DERIVED and ILLUSTRATIVE in capitals in the model's notes and in D.4.2
After. in lower case, as in B.9; the site's assumption chips and the mock-up's read the tag in any case
Why. Fix 8, carried to the notes M7 left: one style for the tag
Before. 'rails' fifteen times, 'bench', 'engine', 'cliff' and 'levers'; 'tiers only go up' and 'centre tier'; supports run together in one 47-word sentence; the page's metadata said version 1.4
After. 'the CDC voucher system', 'schemes', 'a pool of cooks'; 'grades' for Bronze to Platinum and 'centre class' for bands; the six supports as six short sentences; sentences over 30 words split; the version is 1.7, and a test fails if it differs from versions.json
Why. Site-copy audit of 30 October 2026: the site's own words, in step with the papers' 1.7 names and plain style (day.json, day-view.ts and the page's intro)
Before. the subsidised and HDB stall labels were two-sentence blurbs; $1,250 was 'the median rent across NEA's centres'; 'incumbent package', 'PWM', 'parity test', 'Monte Carlo', 'θ' and '$/h' were unglossed
After. 'Subsidised NEA stall' and 'HDB coffeeshop'; $1,250 is 'the median assessed market rent'; 'the plan's package', 'Progressive Wage', 'middle 80% of stalls' and a plain label for the landlord's share
Why. Site-copy audit of 30 October 2026: the site's own words, in step with the papers' 1.7 names and plain style. The labels show on the radio buttons, the table rows and the chart
Before. 'the funding engine', 'browser port', 'parity test', 'the 21-year series', 'the 5,000-future runs'; 'The Technical Paper's series'; abstracts that read 'start here: it reads on its own'; a four-comma film caption
After. 'the funding model for the browser', 'the series for Years 0 to 20', '5,000 simulated futures'; 'Series sources, from The Technical Paper'; short sentences throughout
Why. Site-copy audit of 30 October 2026: the site's own words, in step with the papers' 1.7 names and plain style
Before. the page's description spelt the indicator count from a list that stopped at 'Twenty', so the 21 indicators printed as '21'
After. the count is spelt by a function for any number to 99, so the description begins 'Twenty-one'
Why. Site-copy audit of 30 October 2026: the site's own words, in step with the papers' 1.7 names and plain style. The count comes from data/tracker.json (3 published, 11 in part, 7 not at all)
Before. centres in 'Tier A–C'; 'Lighthouse Centres' and 'Addendum A' and 'Addendum B' as the names of proposal and figure groups; 'rails', 'cliff' and 'second tier'; 'tier' for Bronze to Platinum; 'prep co-operative'; the Plate Tracker as 20 indicators with 10 in part
After. 'Class A–C' centres; 'Year Zero and the Lighthouses and Pilot Centres', 'Hawkers and Social Protection' and 'Follow the Plate' as group names; 'CDC voucher system'; 'grade' for Bronze to Platinum; the old heading of §2.3 replaced by its 1.7 heading; 21 indicators, 11 in part, 7 not at all. Proposal addresses are unchanged
Why. Site-copy audit of 30 October 2026: the site's own words, in step with the papers' 1.7 names and plain style. Strings that the papers quote word for word are not touched
Before. cover subtitles of 52, 50 and 46 words, with a list of seven inside a clause, five semicolons, and 'levers' and 'downside mode' as shorthand
After. the same facts as short sentences; 'the options for the Hawker Protection Floor' for 'the levers'; the register's own description (versions.json 'about') in five sentences
Why. Site-copy audit of 30 October 2026: the site's own words, in step with the papers' 1.7 names and plain style
Before. version 1.2, 'a policy framework', nine PDFs, six companion volumes, $14.99M and 65%, the 'Heritage Watchlist', the page counts of 1.2, no row for G, Sources or Your Table
After. version 1.7 of 2026-10-30, a plan in twelve PDFs, seven companion volumes, $19,683,808 and 72.4%, the Regulars' List, the page counts of 1.7 and a row for G, Sources and Your Table; the 1.5 entry's longest bullets cut into lines. check_publications.py now fails if README.md or CITATION.cff lacks the register's version
Why. Site-copy audit of 30 October 2026: the site's own words, in step with the papers' 1.7 names and plain style. The two files had stayed at 1.2
1.6 30 October 2026 10 changes · shared version
Four calls 1.5 left open, settled on the evidence. NEA already staggers the rent when a subsidised stall is assigned, so the HSS ask narrows to writing that stagger into the Scheme's own terms, and Year 0 asks whether it already applies. The 2012 Panel reported in April 2012. MOM's own rules answer the foreign-worker question, so it is now a recorded fact, not an open one. And Liang Eng Hwa's 2018 question on hawkers managing their own centres gets its credit. The model does not change.
Approved by Uncle James · tag v1.6 · release commit 40c0a28
Before. Give the HSS successor the stagger NEA gives other assignees; under the HSS the successor pays assessed market rent "from the start"
After. NEA already glides the rent where the jump is steepest; the ask narrows to NEA writing its own three-year stagger into the Scheme's terms, to the band at a pilot centre. Year 0 asks whether it already does; if so, the line becomes a credit, not an ask
Why. The HSS FAQs (14 Mar 2022) say only that the successor "will pay assessed market rent to NEA"; "from the start" was ours. For non-subsidised veterans NEA's renewal moves rent towards AMR by no more than $300 a term (4 Nov 2024). Uncle James, 30 Oct 2026, 01:21: the four open calls, settled as the research note of 30 Oct 2026 recommends
Before. The HSS successor "jumps straight" (§1.2) or "jumps immediately" (§2.3) to assessed market rent
After. The successor pays assessed market rent
Why. Follows from the stagger on assignment; NEA is silent on whether HSS hand-overs use it
Before. Year one of the stagger NEA gives an ordinary assignee, "which the plan asks it to give HSS successors"
After. Year one of the stagger NEA gives when a subsidised stall is assigned, "which the plan asks NEA to confirm for HSS successors"; the model is unchanged
Why. Follows §2.3; the modelled stagger (year one of three, to $1,250) stays, marked as an assumption
Before. The Hawker Centres Public Consultation Panel (2011–12)
After. The Panel (November 2011 to April 2012); the Ministry accepted most of its report on 10 May 2012 (The Straits Times, 11 May 2012, as reported)
Why. The Hawker Centre 3.0 Committee's report (Feb 2017, p. 9) and The Straits Times (11 May 2012) both put the final report in April 2012; the Panel's last meeting was on 27 Mar 2012, so February cannot be the final report
Before. The Panel's sources differ on whether it went to the Ministry in February or April 2012
After. It went to the Ministry in April 2012; February is when its recommendations were first shared; the acceptance of 10 May 2012 is known only as reported
Why. The Record's note on what could not be confirmed, brought up to date
Before. Panel report: "chaired by Elim Chew; 18 members; met Nov 2011 – Mar 2012", ¶12–19. The assignment form without a web address
After. The Panel's full citation line (formed Nov 2011; seven meetings to 27 Mar 2012; undated; submitted April 2012, per HC 3.0, p. 9), ¶2 and ¶12–19; The Straits Times, 11 May 2012, added; HC 3.0 p. 9 added; the assignment form's web address and NEA's family-transfer form (28 Aug 2018) added
Why. The sources for the calls above, each in Sources once
Before. The State built hawker centres from the early 1970s "to move street hawkers indoors, for hygiene and food safety, and resettled the last of them by 1986" (uncited)
After. To resettle street hawkers, "so as to improve hygiene and food safety standards" (Masagos Zulkifli, Official Report, 19 Nov 2018), finishing by the mid-1980s (Panel report, ¶2)
Why. A primary source, not a party publication; the Panel says 1985 and others 1986, so "by the mid-1980s"
Before. The social-enterprise row named the 2018 questioners but not the alternative put to the House
After. Liang Eng Hwa (PAP) asked on 19 Nov 2018 whether hawkers could "come together to self-manage" a centre on a strata-title model; Dr Amy Khor replied the Government was "open to any model" that met its social objectives (Official Report, 19 Nov 2018, Vol 94, No 85)
Why. Credit where it is due; dated to 2018, not the 2024 debate, where Liang did not raise it
Before. A chain's stall may draw on its operator's foreign-worker quota, up to 35% of staff (Chua, IPS, 2024; "to be confirmed with MOM"); a question, not a proposal
After. MOM's rules answer it: no work permit holders at any hawker centre, operator-run included (MSE, 21 Apr 2023), LTVP holders since January 2025; in coffeeshops and food courts only the food-shop licensee may hire foreign workers, for its own stalls, at one to two locals (MOM FAQ, 14 Mar 2024; MOM, 25 Feb 2026). Recorded as a fact, not an ask
Why. The question is answered; the Government kept its rules on 10 Sep 2026, so a new ask would repeat it
Before. A chain's stall in the same coffeeshop may draw on its operator's quota (Chua, IPS, 2024; "to be confirmed with MOM")
After. The coffeeshop's or food court's main operator, which holds the food shop licence, can hire foreign workers for the stalls it runs itself (MOM, FAQ, 14 Mar 2024)
Why. As F.9: it is a rule about licences, not brands
1.5 29 October 2026 70 changes · shared version
Three doors, one library. A politician reads the Executive Summary and the Plan, which falls from 72 pages to 44: every proposal now opens with the same card (what it is, who it is for, what it costs, who pays, what it asks of the State, what Year 0 must confirm and where it is proved), and every block the Plan no longer prints is printed in full in a volume, with its old route redirected. An official reads volumes A to G: A becomes The Record (who helps, what was said and done since 2012, the objections answered), and a new volume G, The Technical Paper, prints every dataset, simulation and sensitivity straight from the models, with the downloads and a parity statement. A diner reads Your Table, four pages on the plate, the cook and what a regular can do. And every reference the library cites is now printed once, in a document of its own, Sources: numbered, with where each is used and its web address, so that anyone can check us. The paper colour now runs to the edge of every page. No proposal, figure or claim changes.
Approved by Uncle James · tag v1.5 · release commit 511d6cf
Before. §1.2 in full, with the scheme-by-scheme table
After. §1.2 keeps its intro, the dated record with the panels credited, and the pattern; the full section, table included, is the Record, A.2.1
Why. Destination map: the lessons scheme by scheme go to A
Before. §1.3's lead, the Government's rails table, the intro of §1.3.1, the map and its read-across, the closing of §1.3.2
After. The map (Fig. ES.1), its read-across and a pointer; §1.3 in full is the Record, A.1.1
Why. Destination map: §1.3 to A in full; the Plan keeps the infographic and the half-page read-across
Before. Who Already Helps: A.1 Government's rails (§1.3), A.2 Who else already helps (§1.3.1), A.3 What only this plan does (§1.3.2)
After. The Record: Who Already Helps, and What Was Said. A.1 Who helps today (A.1.1 = §1.3, A.1.2 = §1.3.1, was A.2); A.2 The record since 2012 (A.2.1 = §1.2; §2.1's ownership box); A.3 What Parliament said, sitting by sitting (new table; §2.2's Panel paragraph); A.4 The objections, answered (§1.1's fallacies box and two paragraphs; §2.2's four objections, the fifth and 'Not a subsidy'); A.5 The UNESCO record (§3.6's intro, at-risk count, 'Why UNESCO needs the count' and the Registry; §3.8's associations and 'Two owners, one table'; §5.5's 'What it must not become'); A.6 NHB, fairly (§3.2's 'Not the first record', §4.2's 'Set against what exists', §5.5's 'What it is not'); A.7 What only this plan does (§1.3.2, was A.3)
Why. Uncle James, 29 Oct 2026: A becomes the record of what has been tried; slug who-already-helps and letter A kept, so no route changes; old A.2 → A.1.2, old A.3 → A.7 in the legacy map
Before. Opener 'This Volume at a Glance' for a review of who helps
After. Opener rewritten as 'The Record at a Glance', with the new reader ('If you want to know what has been tried'), bottom line and contents; sources note re-pointed to A.1.2, A.2.1, A.3 and A.7
Why. The volume's new structure
Before. In full, 3,106 words
After. Card, the five band rules and 'Why we think this works'; §2.2 in full is Follow the Plate, D.14 (rents, the newcomer's path, the operator-run premium, the unpublished fees, the fee subsidy); the Panel paragraph and the objections also in the Record, A.3–A.4
Why. Destination map
Before. In full
After. Card only; §2.3 in full is Succession Economics, E.3
Why. Destination map: §2.3 to E
Before. In full
After. Card, lead and At your table; §2.4 in full is Hawkers and Social Protection, C.12
Why. To bring the Plan to 44 pages; C already holds 'whoever the landlord' (C.7)
Before. In full
After. Card, the price-freedom and frame items, the blockquote and At your table; §2.5 in full (incidence, the tranche arithmetic) is Follow the Plate, D.15
Why. Destination map: §2.5 to D
Before. In full
After. Card and the opening paragraph; §2.6 in full is Follow the Plate, D.18
Why. To bring the Plan to 44 pages
Before. In full
After. Card, four of the bench items and At your table; the booking economics and the named deputy are the Budget Book, B.12 'The Relief Cook's Day, Costed'; the lead, the relief notice and the farm models are Hawkers and Social Protection, C.11 'When Illness Closes the Stall'
Why. Destination map: §2.7 to B (costing, the day rate) and C (medical leave)
Before. In full
After. Its first and last paragraphs; the three reasons also printed in Method, Levers and Data, F.9
Why. To bring the Plan to 44 pages
Before. In full
After. Card, its first paragraph and the two covers; §2.9 in full is Hawkers and Social Protection, C.10
Why. Destination map: §2.9 to C
Before. In full
After. Card, its first paragraph and the box 'What we count, and why not the median'; §2.10 in full is Method, Levers and Data, F.6
Why. Destination map: §2.10 to F, the metrics box kept in the Plan
Before. §2.11.1–§2.11.2 (Addendum A's A.7 and A.9) in full, with an 'In one minute' box
After. Card, the lead, the proportionality test, the proposals table and 'What hawkers should not get'; A.9's lead and what stays unasked. Both in full are a new section of Hawkers and Social Protection, C.9 (C.9.1, C.9.2)
Why. The floor's long argument belongs with the comparison it rests on; the Plan from 72 to 44 pages
Before. In full
After. Card, three of the circle items, a credit line to the 2012 Panel and the Hawker Centre 3.0 Committee, and At your table; §2.12 in full is Follow the Plate, D.16
Why. Destination map: the bulk-buying arithmetic to D, the Panel/HC 3.0 credit kept
Before. In full
After. Its intro and a short list of the targets; §3.5 in full is Method, Levers and Data, F.7
Why. Destination map: §3.5 to F
Before. In full, 2,076 words
After. Card, 'a compliment' and 'one way on, and only up'; the tiers, thresholds, what a listed stall gets and the cost are Follow the Plate, D.17 'The Regulars' List, Tier by Tier'; the intro, the at-risk count, why UNESCO needs it and the Registry are the Record, A.5
Why. Destination map: §3.6 to A (the UNESCO record, the count's purpose) and D (the tiers)
Before. In full, 1,441 words
After. Card, its first paragraph, the casting vote and the sunset clause (about 350 words); §3.8 in full is the Budget Book, B.14; the associations and 'Two owners, one table' also in the Record, A.5
Why. Destination map: §3.8 trimmed, the committee history and 'two owners, one table' to A
Before. In full
After. Card, its lead, the Fellowship's terms and two items; §4.2 in full (the stipend's costing) is the Budget Book, B.13; 'Set against what exists' also in the Record, A.6
Why. Destination map: §4.2 to B
Before. In full
After. Cards and the first items of each; §4.3 in full is Succession Economics, E.4 and §4.4 E.5, with §5.5's 'One charge per sale' and two of its items
Why. Destination map: §4.3 and §4.4 to E
Before. In full, 1,759 words
After. A short form naming the three centres and why, and a pointer; the ranking, the conditions, the centres' table and the case for more than one centre are Method, Levers and Data, F.8 'How the Pilot Centres Are Chosen'
Why. Destination map: §6.3 to F, the Plan keeps the three named centres and why
Before. The volumes printed their own sections and levers
After. Each volume ends with a part 'From the Plan, in full': the Plan's sections it owns, word for word, numbered on from its last id so that no existing id moves; a reference to one of them from another volume goes there
Why. One Plan, not two: the mechanism of 1.2 (sections cut from the Plan and placed in the volumes) extended
Before. Each opener's 'In this volume' and 'Where it comes from'
After. Updated for the sections each volume now prints from the Plan
Why. The volumes' new contents
Before. A lever's label named its Plan section through the volume's own rewrite (e.g. 'B.8', 'D.14')
After. The label always names the Plan's section ('the Plan, §2.2')
Why. Since 1.5 most Plan sections have a full copy in a volume, which the label is not about
Before. Every PDF printed the cream paper colour only inside the page's margin box, a tinted block inside a white border
After. Every document: the paper colour now runs to the page edge (@page background in the shared stylesheet; the body transparent); cover and divider pages unchanged
Why. Uncle James, 29 Oct 2026: the white border was ugly
Before. (no technical volume: the parameters, ledgers and simulations were spread across D, E, B and F)
After. New volume G, The Technical Paper: Data and Simulations (46 pages in a scratch build): every parameter of the three models (295, each marked official, derived or assumption), the stall P&L for six stalls, the Monte Carlo distributions, every funding ledger line Years 0–20, the four stacks, the endowment paths, the succession schedules, the Regulars' List and the Year 0 baseline, one-at-a-time sensitivities with a tornado chart, the time series and a reproducibility section with the parity statement; every table generated by src/model/build_addG.py, composed by src/layers/compose_g.py
Why. Uncle James, 29 October 2026: 'a much more technical policy paper with full data set and simulations as a corollary', for the bureaucrat's door. It adds no new assumptions: where a sensitivity needs a range the models lack, the parameter moves ±20%, and the text says so
Before. (the time series lived in research files outside the repository; D.5.4's chart series were typed into build_addB.py)
After. data/curated/series.csv: 666 values in 196 series (469 official, 150 derived, 47 assumption), each with its unit, mark, source, URL and note, from the two research series files; published as g8-series.csv
Why. So G.8 and anyone checking D.5.4 read one curated file with a source on every row
Before. (not published)
After. Dataset /data/g/g1-parameters.csv with its JSON twin: Every parameter of the three models (295 rows, 7 columns; mark: mixed)
Why. Every table in volume G is published as data, generated with the volume, so the arithmetic can be checked outside the PDF
Before. (not published)
After. Dataset /data/g/g1-stall-inputs.csv with its JSON twin: Every input of the six stall archetypes and the prologue's three cases (29 rows, 12 columns; mark: mixed)
Why. Every table in volume G is published as data, generated with the volume, so the arithmetic can be checked outside the PDF
Before. (not published)
After. Dataset /data/g/g2-pnl.csv with its JSON twin: The stall P&L by archetype, a month and a year (19 rows, 14 columns; mark: derived)
Why. Every table in volume G is published as data, generated with the volume, so the arithmetic can be checked outside the PDF
Before. (not published)
After. Dataset /data/g/g2-illness.csv with its JSON twin: What days closed for illness cost the hawker in a year (30 rows, 9 columns; mark: derived)
Why. Every table in volume G is published as data, generated with the volume, so the arithmetic can be checked outside the PDF
Before. (not published)
After. Dataset /data/g/g2-newcomer.csv with its JSON twin: The newcomer's rent path for three bids (6 rows, 13 columns; mark: derived)
Why. Every table in volume G is published as data, generated with the volume, so the arithmetic can be checked outside the PDF
Before. (not published)
After. Dataset /data/g/g3-net.csv with its JSON twin: Net income: deciles, quartiles and the mean, 5,000 simulated stalls a type (7 rows, 16 columns; mark: derived)
Why. Every table in volume G is published as data, generated with the volume, so the arithmetic can be checked outside the PDF
Before. (not published)
After. Dataset /data/g/g3-hours.csv with its JSON twin: Hours: deciles, quartiles and the mean, 5,000 simulated stalls a type (7 rows, 14 columns; mark: derived)
Why. Every table in volume G is published as data, generated with the volume, so the arithmetic can be checked outside the PDF
Before. (not published)
After. Dataset /data/g/g3-rent.csv with its JSON twin: Rent: deciles, quartiles and the mean, 5,000 simulated stalls a type (7 rows, 14 columns; mark: derived)
Why. Every table in volume G is published as data, generated with the volume, so the arithmetic can be checked outside the PDF
Before. (not published)
After. Dataset /data/g/g3-inputs.csv with its JSON twin: The Monte Carlo's inputs: each draw as the code writes it, and its 10th, 50th and 90th percentiles by stall type (17 rows, 20 columns; mark: assumption)
Why. Every table in volume G is published as data, generated with the volume, so the arithmetic can be checked outside the PDF
Before. (not published)
After. Dataset /data/g/g4-ledgers.csv with its JSON twin: Every pilot centre and the national core: every line, every year, 0–20 (long format) (729 rows, 6 columns; mark: derived)
Why. Every table in volume G is published as data, generated with the volume, so the arithmetic can be checked outside the PDF
Before. (not published)
After. Dataset /data/g/g4-stacks.csv with its JSON twin: The four funding stacks, Years 0–3: two, three, four and five pilot centres (16 rows, 16 columns; mark: derived)
Why. Every table in volume G is published as data, generated with the volume, so the arithmetic can be checked outside the PDF
Before. (not published)
After. Dataset /data/g/g4-plan.csv with its JSON twin: The central plan, Years 0–20: every inflow and every use (21 rows, 29 columns; mark: derived)
Why. Every table in volume G is published as data, generated with the volume, so the arithmetic can be checked outside the PDF
Before. (not published)
After. Dataset /data/g/g4-endowment-paths.csv with its JSON twin: The endowment paths, Years 0–20, with the probability of holding real value (84 rows, 12 columns; mark: derived)
Why. Every table in volume G is published as data, generated with the volume, so the arithmetic can be checked outside the PDF
Before. (not published)
After. Dataset /data/g/g4-chua.csv with its JSON twin: Heritage Plate licensees that fail each year: none, a third, half (Chua, 2024) (3 rows, 16 columns; mark: derived)
Why. Every table in volume G is published as data, generated with the volume, so the arithmetic can be checked outside the PDF
Before. (not published)
After. Dataset /data/g/g4-maintenance.csv with its JSON twin: The maintenance rate from Year 4: 10% to 40% of the planning count (4 rows, 6 columns; mark: derived)
Why. Every table in volume G is published as data, generated with the volume, so the arithmetic can be checked outside the PDF
Before. (not published)
After. Dataset /data/g/g5-schedules.csv with its JSON twin: Succession Mortgage schedules under the adopted terms, by stall type, year by year (135 rows, 7 columns; mark: derived)
Why. Every table in volume G is published as data, generated with the volume, so the arithmetic can be checked outside the PDF
Before. (not published)
After. Dataset /data/g/g5-summary.csv with its JSON twin: Succession Mortgage by stall type: goodwill, payment, payoff and royalty (7 rows, 10 columns; mark: derived)
Why. Every table in volume G is published as data, generated with the volume, so the arithmetic can be checked outside the PDF
Before. (not published)
After. Dataset /data/g/g5-royalty.csv with its JSON twin: The Family Name Royalty's ten years and what is left at each anniversary (66 rows, 4 columns; mark: derived)
Why. Every table in volume G is published as data, generated with the volume, so the arithmetic can be checked outside the PDF
Before. (not published)
After. Dataset /data/g/g5-licences.csv with its JSON twin: Heritage Plate licence income and its three shares, by failure rate (24 rows, 6 columns; mark: derived)
Why. Every table in volume G is published as data, generated with the volume, so the arithmetic can be checked outside the PDF
Before. (not published)
After. Dataset /data/g/g6-regulars.csv with its JSON twin: The Regulars' List in orders: every stall and dish type (11 rows, 20 columns; mark: assumption)
Why. Every table in volume G is published as data, generated with the volume, so the arithmetic can be checked outside the PDF
Before. (not published)
After. Dataset /data/g/g6-tiers.csv with its JSON twin: The Regulars' List thresholds (4 rows, 3 columns; mark: assumption)
Why. Every table in volume G is published as data, generated with the volume, so the arithmetic can be checked outside the PDF
Before. (not published)
After. Dataset /data/g/g6-baseline.csv with its JSON twin: The Year 0 stall baseline: sample sizes, interview hours and cost by centre (5 rows, 7 columns; mark: assumption)
Why. Every table in volume G is published as data, generated with the volume, so the arithmetic can be checked outside the PDF
Before. (not published)
After. Dataset /data/g/g6-publishes.csv with its JSON twin: What the baseline publishes, as distributions (7 rows, 3 columns; mark: assumption)
Why. Every table in volume G is published as data, generated with the volume, so the arithmetic can be checked outside the PDF
Before. (not published)
After. Dataset /data/g/g7-sensitivity.csv with its JSON twin: One at a time: every parameter tested, its range and what it does to the four-year total and the endowment (30 rows, 10 columns; mark: derived)
Why. Every table in volume G is published as data, generated with the volume, so the arithmetic can be checked outside the PDF
Before. (not published)
After. Dataset /data/g/g8-rents.csv with its JSON twin: Rents, 2012–2026 (12 rows, 9 columns; mark: mixed)
Why. Every table in volume G is published as data, generated with the volume, so the arithmetic can be checked outside the PDF
Before. (not published)
After. Dataset /data/g/g8-fees.csv with its JSON twin: Fees and wages, 2012–2026 (15 rows, 9 columns; mark: mixed)
Why. Every table in volume G is published as data, generated with the volume, so the arithmetic can be checked outside the PDF
Before. (not published)
After. Dataset /data/g/g8-prices.csv with its JSON twin: Prices, 2012–2026 (15 rows, 13 columns; mark: mixed)
Why. Every table in volume G is published as data, generated with the volume, so the arithmetic can be checked outside the PDF
Before. (not published)
After. Dataset /data/g/g8-sales.csv with its JSON twin: Real sales, 2012–2026 (14 rows, 9 columns; mark: mixed)
Why. Every table in volume G is published as data, generated with the volume, so the arithmetic can be checked outside the PDF
Before. (not published)
After. Dataset /data/g/g9-parity.csv with its JSON twin: Parity between the Python models and the browser ports (1 rows, 8 columns; mark: derived)
Why. Every table in volume G is published as data, generated with the volume, so the arithmetic can be checked outside the PDF
Before. No document written for diners; the library's doors were the Executive Summary, the Plan and six companion volumes
After. Your Table: four pages in Uncle James's voice for people who eat at hawker centres, a site page (/your-table/) first and a four-page PDF second, both built from one source (src/your_table.html), with no ids and the independence disclaimer in the running footer
Why. Uncle James's decision of 29 October 2026 to add a diner's door, so the plan can win over consumer foodies as well as politicians and officials
Before. (new)
After. Where a $4.50 plate at a tendered NEA stall goes ($2.00 ingredients, $0.74 the assistant, $0.37 the landlord, $0.81 the cook's family), and who cooks it: about 60; about $1,470 a month against the riders' $1,800; about 13 hours a day, $9.69 an hour against the assistant's $11.64; $372 a day closed for illness
Why. Follow the Plate's own plate split and the key figures, every number a figure span checked by build_data.py --strict. The model's typical plate is $4.50, not $4, so the page uses Follow the Plate's worked example rather than a new split
Before. (new)
After. The plate over time in four sentences and one chart (price, plates and pay, 2012–2025, 2019 = 100, drawn from the series behind Follow the Plate's "Dearer plates" figure); the median rent of about $1,250 that did not move, the subsidised share falling from 53% to about 30%, and the newcomer at the about $1,800 bid paying $18,900 above a $1,450 valuation
Why. Follow the Plate's account of the plate over time, told to a diner; the chart reads the series from build_addB.py and retypes nothing
Before. (new)
After. The Regulars' List in the diner's terms: one named weekly regular, one order counted once; 20 names list a stall; Silver, Gold and Platinum at 50, 100 and 250 names or the share route; tiers only go up, no way down; a durable base of about a third of orders (35%, an assumption), about 168 regulars at a typical stall; a compliment, never a grade
Why. The Plan's Regulars' List and Follow the Plate's tiers by orders, unchanged in substance, said to the people who would sign it
Before. (new)
After. Five things a diner can do (endorse a dish, look for the Heritage Plate mark, spend vouchers at a stall and back the hawker tag, give to the Endowment with the $10M match, visit a Lighthouse or walk through A Day at the Lighthouse) and the one ask of the Government: count the stalls at risk and put the count in the next UNESCO report, due on 30 June 2027
Why. The Plan's ways in for the public, its ask on the count and the UNESCO deadline, gathered on one page
Before. (new)
After. Figure spans: every number on the four pages carries a key-figure id, a model expression or a verbatim quotation of the canonical text, checked by compose_t.check_spans from build_data.py --strict and by the site test; Document Control prints as a colophon on page four (the history is in the Change Log)
Why. The brief's rule that every number comes from key-figures or the model and is checked, applied to a document that prints no ids
Before. (new) The library's references were printed in Method, Levers and Data (F.5) as three lists one after another, the Plan's, then those of Hawkers and Social Protection, then those of the Follow-the-Plate model; a reference cited in more than one place was printed in more than one list
After. Sources: every reference in the library once, 455 in all, numbered S.1–S.455 and grouped by the document that cites it first (the Plan, S.1–S.238; Hawkers and Social Protection, S.239–S.323; Follow the Plate, S.324–S.425; the Technical Paper's series, S.426–S.455). Each entry gives the reference as the library's list printed it (or its body and title), its date, where it is used, in the ids readers see, and its web address (421 have one), or "No URL; supplied as a document"; then an index by body, a note on what the list does not say, the independence disclaimer and Document Control. Composed by src/layers/compose_s.py, never typed
Why. Uncle James, 29 October 2026, yes to a standalone Sources document, so that anyone can check any citation in one place
Before. data/curated/sources.json had 286 records; about 170 references the lists printed (most added in 1.3 and 1.4, for who already helps and the record) had none, so the site's Sources page left them out
After. 455 records. The 169 missing were generated from the list items as printed (body, title, date, web address, and the list and heading that print them); every record a list prints keeps the list's own wording (as_listed), which Sources prints. build_data.py --strict fails if Sources prints a record twice, prints one with no record or leaves one out; check_layers.py fails unless every item and web address of the canonical lists is in Sources
Why. Sources must be complete before it can replace F.5, and the PDF and the site must number the same list
Before. F.5 References: the Plan's references, then the sources and notes of Hawkers and Social Protection, then those of the Follow-the-Plate model, about eleven pages
After. F.5 References: Now in Sources, a half-page note after F.4: the references for the whole library are in Sources, S.1–S.455; the data and downloads are F.4. F.5 keeps its anchor; the legacy map records the move (library: F.5) and the 1.2–1.4 route of the Plan's references, /method-levers-data/references-and-context/, redirects to /sources/. The cover, the opener and the register's description name Sources
Why. One list, not three; the volume falls from 40 pages to 30
Before. The full source list of Hawkers and Social Protection, printed again in F.5
After. The 116 references the volume uses, in short (body, title, date and where in C), each with its S-number, in three columns after the volume's sources note; the list's opening line and its open questions stay, under its own title
Why. The full entry, with its web address, is in Sources; the volume's list says where to find it. C stays at 32 pages
Before. The full source list of the Follow-the-Plate model, printed again in F.5
After. The 155 references the volume uses, in short (body, title, date and where in D), each with its S-number, in three columns, under the list's opening line
Why. As for Hawkers and Social Protection; D falls from 54 pages to 52
Before. The 100 source strings of data/curated/series.csv, as the file names them, each with a link; G.1 and the list sent the reader to F.5 for the full citations
After. The 95 sources behind the series in short, each with its S-number (the file's five status notes, such as "None, an assumption", stay in its own source column); G.1 sends the reader to Sources
Why. As for the volumes; G stays at 46 pages
Before. Each document's sources note: "The references for the whole library are collected in Method, Levers and Data (F.5)"; F.4's layers timeline gave "the sources of D.5.4 in F.5"
After. Each names Sources instead: the Plan's note, "the references for the whole library are in Sources … (S.1–S.238 for those the Plan cites first)"; C and D say their lists are in short with the S-numbers; F.4 says "listed in Sources"
Why. F.5 no longer holds the lists
Before. /sources/ listed the 286 records in curated order, citing them by their canonical ids; the Plan's references had a page of their own in Method, Levers and Data
After. /sources/ shows all 455 in the PDF's order and S-numbers, grouped as the PDF groups them, each with where it is used linked to the page that uses it, and keeps every record's old anchor; the Plan's references page is served by /sources/ (301s from /framework/, /main-paper/ and /method-levers-data/references-and-context/); the downloads page lists the Sources PDF with the library, now ten documents
Why. The site and the PDF read one list with one numbering
1.4 29 October 2026 39 changes · shared version
The library now reads the Government's own record first (the 2012 Hawker Centres Public Consultation Panel, HC 3.0, the Official Reports of 12 March 2013 and 19 November 2018, Singapore's UNESCO periodic report, the ACPW report and Chua's IPS paper) and answers the objections it raises: hawker pay has been measured once, about $1,470 a month in 2021, and is reconciled with the model; the health evidence enters C and §2.9, where a day shut for illness costs a typical tendered stall about $372; ten reassurances are set against what they leave out, the count is kept as distributions, not medians, and one standing NEA–NHB–FMAS body is asked to own it. The Regulars' List counts orders, not plates, a Year 0 stall baseline sets its thresholds, and the four-year budget at three centres is $19,683,808. Granted, the record also cuts the plan down to size: on Chua's evidence a third of Heritage Plate licensees are assumed to fail each year, and the Government's match now covers part of the base (40%), not most of it.
Approved by Uncle James · tag v1.4 · release commit be677f3
Before. A regular is one named diner who eats at the stall at least weekly; Silver, Gold and Platinum at 50, 100 and 250 endorsements or 8%, 15% and 40% of the stall's weekly plates, at 1.5 visits a week
After. A regular is one named person who orders from the stall at least weekly; the share route counts orders, not plates: 10%, 20% and 50% of the stall's weekly orders, at 1.5 orders a week each; the counts route (20, 50, 100, 250 names) unchanged
Why. Uncle James, 29 October 2026: a family's takeaway is one order of several plates, so it lifts the plate count without lifting the count of regulars. The shares are set just below what 50, 100 and 250 regulars order at the typical tendered stall (10.4%, 20.8%, 52.1% of 720 orders a week), so it reaches each tier at about the same number of regulars as before
Before. (no plates an order; one plate a visit)
After. Plates an order by dish type, ASSUMPTION until the baseline: about 1.2 for drinks and snacks, 1.3 for noodles and rice, 2.0 for family dishes (kway chap, zi char, roast meats); dollars an order = the plate's price × plates an order; orders a week derived from plates a week
Why. No published Singapore figure gives plates or items an order at hawker stalls; the nearest, Hawkers Go Digital's $45 million in 5.4 million e-payment transactions in August 2023 (Enterprise Singapore, 26 Oct 2023), is about $8.33 a transaction, about 1.9 of the model's $4.50 plates, a ceiling since small orders are more often paid in cash
Before. The healthy ratio: 35% of a stall's plates from its weekly regulars, about 219 regulars at the typical tendered stall; John reaches Silver with 34
After. 35% of a stall's orders from its weekly regulars, with the same sources (Toast and Resy count orders), about 168 regulars at the typical tendered stall; John, cooking kway chap at 2.0 plates an order, reaches Silver with 21
Why. The healthy ratio restated in the new unit; John and Ah Kow cook kway chap, a family dish
Before. The Regulars' List: What a Tier Means in Plates; one table by stall case
After. The Regulars' List: What a Tier Means in Orders; a table by stall case (dish type, plates an order, orders a week, $ an order, regulars for each tier) and a table by dish type at the typical stall's plates; a paragraph saying every number is an assumption until the baseline
Why. Uncle James asked for the tiers by dish type as well as by stall case
Before. (no stall-level baseline; the evaluator's baselines only)
After. A stall baseline in each pilot centre's year before it opens: a two-week till count at every consenting stall (orders a day, plates an order, dollars an order, the share of orders from regulars the hawker names), by dish type, gathered by the listening teams through oral interviews and a tally sheet, published only by dish type and centre; it sets plates an order and the share thresholds by dish type before the first listing, then gives the Plate Ledger its first stall-level numbers
Why. Uncle James, 29 October 2026: collect baseline data from oral interviews to re-establish plate counts by stall and dish type and dollars an order
Before. Cost shares: publish NEA's survey yearly; a voluntary panel of about 300 stalls for what it misses
After. … seeded by the Year 0 stall baseline (§6.1); the Plate Ledger box says its first stall-level numbers come from the baseline
Why. The baseline's second use; data/tracker.json kept in step
Before. (no note on the baseline's or plates-an-order assumptions)
After. The stall baseline's interviewer hours and rate, plates an order by dish type and the order rate are named as our assumptions, with Hawkers Go Digital's totals as a ceiling only
Why. Every unverifiable claim is marked in its volume's sources note
Before. No official source measures what hawkers earn
After. One official figure: about $1,470 a month in 2021, the median for full-time own-account hawkers (MOM CLFS 2021, in the ACPW report, Fig. 1.1, read from a chart); a box reconciles it with the model's $3,169 for the typical tendered stall in 2026 (2021 prices, own-account only, the median person, a survey), marked derived
Why. The Advisory Committee on Platform Workers published the figure in 2022; set beside the model, it sits in the model's lower tail: of all simulated stalls pooled, 29% net less than $1,760, its value in 2026 prices
Before. (the newcomer's rent shown only as the 2023 median bid)
After. The newcomer's path under NEA's November 2024 renewal rule: a $1,800 bid against a $1,450 valuation pays $1,800, $1,625 and then $1,450, about $18,900 above the valuation over six years ($29,700 against $1,250); $8.01, $8.54 and $9.08 an hour on the typical stall; KF Seetoh's 'six years' stated fairly. plate_model.newcomer_rent_path, ported to the simulator
Why. The Online Citizen, 15 Sep 2026, on MSE's reply of 8 Sep 2026: the rule takes the entrant to the valuation only from Year 7; the valuation is illustrative because NEA does not publish it
Before. (no cost of a closed stall)
After. A day shut for illness costs the typical tendered stall's hawker about $372 ($342 subsidised, $441 HDB-rented coffeeshop); six weeks about $13,400, 35% of the year's net; a rent waiver saves about $48 a day; a Corps cook on medical leave costs $40. plate_model.illness_cost, ported to the simulator
Why. Worker H2's arithmetic for C and §2.9, checked against the model: rent, fees, the assistant, utilities and other running costs continue while the stall is shut
Before. About half of the cooked-food stalls on rent paid subsidised rents in 2012–13; the roughly 600 cooked-food stalls sold in 1994–97 came back onto rent in 2014–17
After. 53% of cooked-food hawkers on subsidised rent in March 2013, about 40% in 2020, about 30% in 2023; about 2,000 hawkers at 15 centres saw their Stall Ownership Scheme leases end in 2014–17, with Grace Fu's three options; about 600 of the stalls sold were cooked food
Why. Dr Balakrishnan and Grace Fu, Committee of Supply, 12 Mar 2013 (Official Report, Vol 90, Issue 13); MTI, 2015
Before. Subsidised stall, original terms (rent glides to the $1,150 band): successor after payment $1,784, after royalty $1,378
After. Two rows: a successor from outside the family, who steps to the $1,250 median assessed market rent on NEA's three-year stagger ($1,751; $1,345), and a family successor, who keeps the subsidised rent ($2,071; $1,665); the goodwill sum ($50,538) is the same in both
Why. Dr Koh, 13 Nov 2024: the pioneer rent 'can be transferred to the immediate family members … at the same low rent'; NEA staggers market rent over three years for assignees (assignment form, 11 Dec 2020) but not, today, for HSS successors (HSS FAQs, 14 Mar 2022)
Before. Heritage Plate licence income with no failures: about $90K in Year 3 and $3.0M a year from Year 7
After. A third of licensed outlets and packaged lines assumed to fail and pay nothing in any year (ASSUMPTION; tested at none and half): about $60K in Year 3 and $2.0M a year from Year 7; E.1.1 explains why the royalty is on takings
Why. Nigel Chua (IPS, June 2024, pp. 31–48): most licensed or acquired hawker expansions shrank or closed, and the one profit-share licence, Lao Chaozhou's, paid the founders nothing
Before. Twenty indicators; income and hours missing; rents as medians; floor costs as four medians
After. Twenty-one: income and hours as distributions (quartiles, the mean, the bottom tenth) by stall and centre type, partly published (the 2021 median); a new 'Health and cover' line; 'Flows: entries, exits and transfers'; rents, bids and the newcomer's path as distributions; every floor charge on its own line (MSE's three published charges come to 1.5–1.6 times the rent; dishwashing, point-of-sale and fines unpublished); the Regulars' List by orders and regulars' share; the Plate Ledger reports distributions, not medians
Why. Uncle James, 29 Oct 2026, 'metrics, not medians'; data/tracker.json kept in step
Before. "No data" / "No official source measures hawkers' earnings"; C.5 Income: "No official figure"
After. About $1,470 a month in 2021, median gross profit of full-time own-account hawkers (MOM CLFS 2021, ACPW report Fig. 1.1, read from a chart), below P20 ($2,200), riders ($1,800) and private-hire drivers ($2,000); caveats: 2021, own-account only, before tax
Why. Correction 2: MOM has measured hawker income once, in the Advisory Committee on Platform Workers' report (17 Nov 2022)
Before. §1.2 opened on the table; the "Not new, and said so" list began with the PSP's ballot
After. A dated history: resettlement from the early 1970s; the 2011 restart ("a ballast"); the 2011–12 Public Consultation Panel (Elim Chew; ¶12–19) and the Hawker Centre 3.0 Committee (2016–17; eleven recommendations accepted 8 Mar 2017; Rec 5, Rec 9; no hawker on it; it backed operator-run centres); A.3 leads with both panels
Why. Addition A: credit the Government's own panels before any party
Before. FMAS "co-chaired the UNESCO nomination"
After. The 14-member nomination committee of 2018–20 was co-chaired by the CEOs of NHB and NEA and FMAS's President
Why. Correction 9 (Grace Fu, PQ reply, 15 Jan 2019; Malay Mail, 23 Oct 2018)
Before. NHB Heritage Grants: "more than 550 recipients since 2013, a figure we could not confirm"
After. More than $16.6M to more than 580 recipients (NHB Annual Report 2024/25); two food houses among the 2024 Stewards; the Organisation Transformation Grant funds heritage food businesses; the Heritage Apprenticeship pilot (Aug 2024); the next Heritage Plan being drafted
Why. Addition E: NHB, fairly
Before. NHB's inventory "does not say which stalls or dishes are at risk"
After. NHB reviews its 106 elements once a year in phases; ICH-050 last revised March 2019; the Registry offered as research for its next revision
Why. Addition D; V corrects Q's 'promise broken' framing
Before. Hong Kong: "No new licences since 1956, transfer only to a spouse"; INMA; Penang's "2024 draft would extend the bar to 13 named dishes"
After. No new hawker licences since 1972 (LCQ3, 2017); dai pai dong successions to sons approved in 2021 and 2023 (LCQ21, 2024); INMA renamed Institut pour les Savoir-Faire Français (21 Mar 2024); Penang's 13-dish extension not in force at Jan 2026
Why. Correction 8
Before. "Apart from the Foochow fund of 1956, we found no hardship fund"
After. Adds the hardship stall-allocation scheme that ended in 1990, which the Government declined to revive in 2021 (MSE, COS, 4 Mar 2021)
Why. U finding 3.2
Before. §2.9 had no health evidence; C.5's "The body is the business" one paragraph
After. §2.9 "Why cover comes first"; C.5 "What the work does to the body" and "Why cover is a heritage measure": WHO/ILO long hours, Makki 2026, Tüchsen 2000, NUS stall fumes, IARC 2A, Leong 2024, Tang 2010, two dated closures; $372 a closed day ($342, $441), six weeks $13,400 (35%), a rent waiver $48 a day, a Corps cook $40; only the Corps reaches the commonest harm
Why. Addition K (H2's health findings; A4's illness_cost)
Before. Kay Lee's "move to a central kitchen"; a subsidised hand-over never priced on a subsidy "the successor will not inherit"
After. The central kitchen and ten outlets never happened and Kay Lee ended as one shop (Chua, IPS, 2024); a child keeps the pioneer rent and needs no mortgage; Balakrishnan's "the key principle is personal operation of the stall" (2013)
Why. Corrections 1 and 6; addition J
Before. Stewards "mostly to the performing arts and crafts"
After. Two food houses among the 2024 stewards; no hawker yet; NHB's Heritage Apprenticeship as the Fellowship's kin
Why. Addition E
Before. Licence income $90,000 (Y3) → $3.0M (Y7); corpus real value 76 of 100; match shortfall $11.2M / $28.9M; corpus 105% / 129% / 86%; $126K carried, short from Year 6; $33.5M without HDB; at 20% no year short, 90%; at 30% $5.8M; at 40% $14.9M; Government licence match $2.0M
After. $60,000 → $2.0M, net of a third of licensees failing; 71 of 100 (20 short); $14.7M / $32.4M (§7.6.3 table from A4); 101% / 121% / 86%; $84K, short from Year 5; $37.0M; at 20% $0.06M short in one year, 87%; $6.9M; $16.0M; $1.4M
Why. numbers_A4.md §6: Chua's failure rate (hp_failure = 1/3, ASSUMPTION); levers and the Budget Book opener follow (one future in five)
Before. C.2 argued from the ILO's dependent contractor only
After. Box "The Act's own test": the ACPW's three management-control factors (¶3.6.2), street-hail taxis left out (¶3.6.4), clientele (¶3.2.8); an NEA-centre hawker is not a platform worker; operator-run centres closest; argued by analogy; Khor's "voluntarily and knowingly" against MOM's own employment factors (19 Nov 2018); the 2018 5-day/8-hour ceiling; Yishun Park contract and the Timbre (Jul 2026) and One Punggol (Aug 2026) exits
Why. Addition C (P, Q, U, V findings)
Before. (hawkers counted as stalls only)
After. 11,400 hawker/stall-holder SEPs in 2016 (TWG Annex B); 11,700 own-account in 2017 (MOM)
Why. Correction 5
Before. The TWG "recommended two things: contribute-as-you-earn, and the use of licensing controls to secure payment"
After. Licensing controls were about insurance (¶45); licence refusal for arrears is a sanction it described (¶48); fn 18 leaves out "individuals who purchase food from hawker stalls"; ACPW Rec 9 as the cure; ¶5.2.6 diagnoses the late bill; the Act's promised review covers only other platform services
Why. Correction 5; addition C
Before. Work injury: employee "WICA, all employees"; hawker "None"
After. WICA limits for accidents from 1 Nov 2025 ($53,000 medical; full wages 14/60 days; death $91,000–$269,000); WICA excludes the self-employed
Why. H2 correction: the ACPW quotes the older limits
Before. Age: "new entrants median 46" (unsourced); Hours: "About 13 a day across 21 interviewed hawkers"
After. 46 cited to NHB and MCCY, 17 Nov 2020, Annex E, with 59 (2016) and 43 (SEHCs, 2018); "worked 13 hours on average", read as a day; Koh's "often more than 14 hours a day"
Why. Correction 3; H2 note 1
Before. (no measured analogue)
After. Riders' CPF adequacy (about 10% of an employee's CPF; just over one in four above 55 at the BRS; ACPW ¶5.2.4–5.2.5), marked as platform data and an assumption for hawkers
Why. Addition C
Before. Proposal 1 precedent: the TWG; proposal 3: "To be priced"
After. Proposal 1 on ACPW Rec 9, kept opt-in on purpose (cash takings; the licence already enforces); proposal 3 priced on WICA's terms (Rec 6), with Malaysia's compulsory PERKESO cover since 1 Jan 2025 (38,358 hawkers contributing, Feb 2026); the ACPW's rejection of fixed-payout group cover (¶4.10.3) answered
Why. Addition C; H2 §3
Before. Co-contributions as a departure "from the principle of self-help"; three reasons; reason 3 "The match is the share that gatekeeper would pay"
After. Quoted as "departs from the principle of self-help", with the TWG's reason (¶54); reason 3 argued by analogy, conceding the Act's test; a fourth reason: the ACPW's co-contribution "part of the unique social compact in Singapore" (2022, accepted)
Why. Correction 5; addition C
Before. Livelihood dataset: earnings, hours, savings and injury
After. As distributions, with musculoskeletal pain (Nordic questionnaire), days closed for illness and the stated reason for every exit
Why. H2 note 4
Before. Rent cap declined "in 2018 and 2024" (unattributed)
After. 2018 (Dr Amy Khor) and 2024 (Dr Koh Poh Koon); brand stalls and operators' quotas noted
Why. Correction 8
Before. 19 Nov 2018 cited from the SG Parliament mirror (debate 9883); the CNA pieces undated; Dorsey (2025), Koh, T. (2022), Tan, K. P. (2021), NEA (2021; 2024) listed; "MAFF (2014), Ita-mae certification framework"
After. Official Report, Vol 94, No 85, pp. 13–35, Oral Answers; 2013 COS as Vol 90, Issue 13; CNA 19 Nov 2018 and CNA/TODAY 20 Oct 2018 dated; the four uncited entries removed; MAFF's 2016 cooking-skills guidelines (3,976 certified by Mar 2026); new reference groups for the record and for health
Why. Corrections 8 and 10
Before. 2014–17: "The ownership leases run out"; subsidised share "from about half … to about 30%"; hardship scheme "1980 or 1989"
After. Fu's three options at expiry; 53% (2013), about 40% (2020), about 30% (2023); 1980, 1989 or 1990 (MSE 2021)
Why. Correction 4; numbers_A4.md §4
Before. Sources notes: the 46 unsourced; Penang and the 13-hour page "would not open"; the 2018 sitting from a mirror
After. Updated: the 46 sourced; the $1,470 read from a chart; health evidence mostly foreign; the 2018 and 2013 sittings from the Official Report; Hack's page opens ("13 hours on average"); F notes the March 2010 record is still sought and that debate 20156 is not it
Why. Corrections 3, 10; addition J
Before. Bottom line (C.1 plaque 5): "The state is not counting"; opener: five points
After. "The state is barely counting" (income once, $1,470); the opener adds the pay point and the health line, a management-control term, and notes the case is by analogy
Why. Corrections 2; additions C and K
1.3 29 October 2026 126 changes · shared version
A grounding release. The library's claims were checked against their sources (314 checked; corrections made where the evidence differed, and what could not be confirmed is named in each volume), and the plan changes where the evidence says it should: the fee reset becomes a fee subsidy, the Hawker Apprentice Work Pass moves to Ideas for Later, apprentices are paid the Progressive Wage for the full 18 months, the Relief Corps is built on NEA's Joint Operator Scheme with a relief day the Employment Act allows, and Chinatown Complex starts after its pending repairs. Follow the Plate gains the middlemen (D.5.3), the plate over time and what a Regulars' List tier means in plates (D.4.2). On Uncle James's decisions of 29 October 2026: the Government gains one board seat with a casting vote, every use of it published within 30 days, and the Hawker Council's vetoes are internal to the Seed Bank, with a voting seat for FMAS; the Master Fellowship pays less each month for longer, a teaching year and a two-year stewardship ($90,750 a Fellow over three years); the Heritage Watchlist becomes the Regulars' List, one way on and no way down, in four tiers, with the at-risk count published only in aggregate; pilot centres are ranked by their likely count of heritage-worthy stalls; Hawkers Succession Scheme 2.0 becomes an extension of NEA's Scheme, and the endowment is the Hawker Endowment; the Family Name Royalty is capped at ten years. The four-year budget at three centres rises from $14,990,340 to $19,581,112.
Approved by Uncle James · tag v1.3 · release commit e9ba5e7
Before. "close to half of tenderers would see their rent much higher than it is today"; "the end of the $1 rental bid stalls"; "only about 4% of stalls pay above the median assessed market rent of $1,250"
After. "close to about half of tenderers will see their rent much higher than it is today"; "the end of the $1 bid stalls that we have in the system today"; "only 4% of the 7,000 cooked-food stalls pay rent above the assessed market rent of $1,250"
Why. Quotations matched to the Hansard record of 13 November 2024 (Parliament 14, Vol 95, Sitting 147), read through Open Hansard; the 1.2 wording followed MSE's copy of the speech. Hansard also confirms Dr Koh's support for Edward Chia's amendment and the other quotations used.
Before. Hansard would not open, so the debate of 13 November 2024 is cited from the ministry's and the party's own published speeches
After. Dr Koh's speech is quoted from the Hansard record as reproduced by Open Hansard, an unofficial site; the amendment's full text and any division remain unconfirmed
Why. Uncle James found the Hansard record on Open Hansard (29 October 2026).
Before. calling on the Government "to continue its support for hawkers by regularly reviewing its policies"; the amendment's text and any division unconfirmed
After. the resolution quoted from the Official Report, including "while enabling hawkers to earn a fair livelihood"; no Member recorded dissent
Why. The Official Report of 13 November 2024 (via SG Parliament) gives the motion as moved by Leong Mun Wai, Edward Chia's three amendments, the resolution as agreed, and the Speaker's call for dissent, which no Member recorded; Leong's closing speech says the PSP would support the amended motion.
Before. The 2024 proposals on the cheap plate, an academy, a foreign hand and the tender were credited to the PSP (and Leon Perera's 2021 motion) only
After. Louis Chua's speech of 13 November 2024 is credited too: Government-funded CHAS discounts, hawkers setting their own prices, a price-quality tender and a rent cap, a single independent academy, and a work pass for hawkers aged 60 and above
Why. Even-handed attribution: a cold political read found the WP's matching proposals uncredited (wp.sg, 'Growing Singapore's Hawker Culture', 13 Nov 2024).
Before. Riders and drivers get platform-paid CPF; 'most riders' have someone paying into their CPF
After. Every rider has work-injury cover; only riders born in or after 1995, and the 27% of older riders who opted in (about 26,500), have an operator CPF share; the 90,000 a month are mostly MediSave-only
Why. W2 #42, #43, #44 (MOM, 12 Jan and 5 Aug 2026)
Before. A MediSave-only hawker reaches 55 with no Retirement Account savings; CPF LIFE pays $0
After. Little or nothing: only MediSave above the Basic Healthcare Sum ($79,000 in 2026) flows into retirement savings, so for most such hawkers CPF LIFE pays little or nothing
Why. W2 #23 (MOM, 7 May 2022; MOM, 26 Jan 2026)
Before. MediSave paid in one annual lump, up to two years after earning; a quarter of the self-employed pay late and lose Workfare
After. Billed once a year after the tax assessment, payable by monthly GIRO; nothing collected as they earn; Workfare lost only if unpaid by CPF's deadline
Why. W2 #18, #21 (CPF, 5 Jun and 22 Oct 2026)
Before. Nothing enforces MediSave; a quarter of stallholders in arrears
After. The SFA stall licence is not renewed with MediSave arrears unless the hawker is on GIRO; the quarter is for all self-employed persons, and hawkers' own rate is probably lower and unpublished
Why. W2 #19, #20 (SFA; CPF, 4 Nov 2025)
Before. No income-loss support when illness shuts the stall
After. No injury cover, sick pay or income-loss insurance of their own; means-tested ComCare Short-to-Medium-Term Assistance and private Prolonged Medical Leave policies (since 2018) exist
Why. W2 #33 (MSF; NTUC, 7 Sep 2018)
Before. Contribute-as-you-earn: no plans to extend it to the private sector
After. Adds that private buyers may adopt it voluntarily, and that the 2018 Tripartite Workgroup asked for both contribute-as-you-earn and licensing controls; for hawkers the licence half is done
Why. W2 #36, #19 (MOM, 22 Feb 2018)
Before. Proposal 1, MediSave as you earn, closes 'the annual lump-sum bill, and the arrears that cut off Workfare'; 'the annual bill and its arrears disappear'
After. Re-framed around smoothing cash flow: the bill is paid in small sums as the takings come in; the licence already enforces payment, so fewer arrears are no longer claimed
Why. W2 new evidence on proposal 1
Before. MRSS precedent stated without scale; the 2018 Workgroup cited only for daily cash
After. MRSS: $456M to more than 250,000 members for 2025, and persons with disabilities under 55 from 2026; the Workgroup's objection that Government co-contributions 'depart from the principle of self-help' is stated and answered in C.8.1
Why. W2 #26, #38
Before. Pooled cover presented as if nothing exists; the Workgroup's daily-cash recommendation as precedent
After. Pooled cover is additive to ComCare and to the medical-leave policies insurers already sell; the pilot asks why so few hawkers buy them; the $750 premium marked an assumption
Why. W2 new evidence on proposal 3; W4 #59, W5 #46
Before. A relief cook covering a hospitalised hawker breaches both rules; the nominee route alone
After. Names NEA's Joint Operator Scheme (since 2012) and the nominee route for illness (Dr Koh, 10 Mar 2025); the Corps is built on them and asks for speed and a short-cover variant
Why. W5 #37, #38, #36
Before. The 2024 extension of Hawkers Go Digital covers government-owned premises only
After. The MDR subsidy, last extended in October 2024 to end-2025, covered government-owned premises (HDB coffeeshops included); no extension into 2026 found
Why. W2 #9
Before. About 805 coffeeshops (about 405 rented from HDB, about 400 sold); about 6,000 coffeeshop stalls
After. 776 coffeeshops in HDB estates: 374 rented from HDB and 402 sold (July 2024); about 5,800 coffeeshop stalls
Why. W1 #36
Before. Coffeeshop stall asking rents have a median near $3,400; NEA rents of about $1,200–1,800, the 2025 median rent paid $1,450
After. Commercial guides put HDB coffeeshop stall rents at about $4,000–10,000 (2026); NEA's median about $1,250, $1,450 the 2025 median at centres comparable with the operator-run ones, $1,800 the 2023 median winning bid
Why. W1 #35, #5; W4 #11
Before. Delivery apps take commissions of up to about 30%
After. GrabFood publishes a 30% service fee, plus GST; Deliveroo left Singapore on 4 March 2026
Why. W4 #30
Before. Smaller wording: Jobseeker Support quote, Silver Support's 5-room rule, Earn and Save Bonus exclusions, top-up tax relief, the Budget 2026 top-up band, the MCCY reply, the median-age source
After. MOM's own words quoted (7 Jul 2026); 'owning' a 5-room flat; AV $25,000 and one-property exclusion; $8,000 own plus $8,000 family, none on matched top-ups; $500 band limited to savings below $60,000; 'written reply (PQ 3030)'; median age dated Oct 2021 and Nov 2024
Why. W2 #35, #30, #31, #28, #29, #1, #54
Before. Sources opened 25–27 October 2026
After. Re-checked on 28 October 2026; twenty new sources (CPF billing and GIRO, SFA licence renewal, the 2018 Workgroup, PML insurance, ComCare, the Joint Operator Scheme, Grab's fee, MRSS scale, the Basic Healthcare Sum, 776 coffeeshops, 2025 medians and others)
Why. Verification pass of 28 October 2026
Before. Sources notes name where the sources are listed
After. Each sources note adds 'What we could not confirm': the ledgers' unverifiable and unreachable claims for that volume, named plainly
Why. Decision 7: every unverifiable claim softened or named in that volume's sources note
Before. Opener: riders 'platform-paid CPF'; 'no pot to retire on', 'CPF LIFE pays him $0'; 'one yearly sum'; 'no income-loss support'
After. Opener restated as in C.1: younger and opted-in riders; little or nothing above the Basic Healthcare Sum; billed yearly, GIRO and the licence; ComCare or a policy he buys; Basic Healthcare Sum added to the terms
Why. W2 #18, #19, #23, #33, #43
Before. No one pays a retiring hawker for his goodwill ... keeps his recipes well enough to cook from, gives a new cook an income floor ..., gives any hawker a day off
After. No scheme pays ... – the few private sales are one-off deals – ... keeps his recipes in a standing archive good enough to cook from, tops up a new cook's own income for two years, sends a relief cook so a hawker can take a day off without closing
Why. W3 #42, #44, #49, #32
Before. Levers: three NEA rents 'about $1,200 assessed, $1,450 paid'; 'the fee's cost' as income the fee collector gives up, the operator bearing its share; delivery 15% instead of 22%; the Corps and named deputy without the Joint Operator Scheme; the matching pool without CMF terms
After. $1,250 the system median and the median AMR, $1,450 the comparable-centre median, $1,800 the winning bid; the fee subsidy, a Government cost against a contractor's price, the 6 Jul 2021 reply cited, a pass-through at operator-run centres; 15% instead of 30%; the circles' saving shared between supplier and cook; the tied-supplier ask; the Corps on the JOS, the deputy a short-cover joint operator; CMF caps and no double matching; MediSave as you earn a cash-flow lever
Why. Decisions 1 and 6; W1 #5; W3 #24; W4 #30; W5 #9, #37, #38, #61; W6 #28
Before. No answer to the 2018 Tripartite Workgroup's objection to Government co-contributions
After. 'The self-help objection': the match pays only on savings the hawker makes first; MRSS already does this at scale; since 2018 the rider has gained a counterparty and the hawker's gatekeeper is the State; the Workgroup's point on mandatory intermediary contributions conceded
Why. W2 #38 and new evidence on proposal 2
Before. Openers: the ask without the fee subsidy; the ledger total without comment; 'no one pays ... gives any hawker a day off'
After. Budget Book: the fee is now a subsidy and the Work Pass has left the ask; the bill is higher because of apprentice pay and the Fellows' CPF; Chinatown's start after its repairs. Who Already Helps: 'no scheme', the private sales, the standing archive, a day off without closing, Kopitiam's HawkerBoss
Why. Decisions 1–4; W3 #32, #42, #44, #49
Before. Levers and openers at the 1.2 model: break point $720; Corps leverage 2.65; $9.42 an hour; $8.0M stretch; $3.50M a year from Year 4; circles $435 a month
After. At the 1.3 model: $610; 2.17; $9.69 (with the package $15.60; 66% → 26% below the benchmark); $7.9M; $4.05M; $407; bench sizes from the model (15→23, 4→6, 8→12)
Why. Model inputs corrected (rent $1,250, delivery 30%, wage shock with CPF, Corps at the PWM cook rate): W1 #5, W4 #11, #30, #31, #49; numbers_M.md
Before. Opener: 'Suppliers take the most'; no word on middlemen
After. 'Ingredients take the most'; the middlemen have not widened their cut (shop food prices tracked import prices, 2019–2025); the real tolls are delivery platforms, tied suppliers and past cartel conduct (D.5.3); the cover and contents name D.5.3
Why. Decision 6
Before. Plate Tracker data: seventeen indicators
After. data/tracker.json: eighteen, with t18 'The margin spread' (in step with D.8); counts 3 published, 9 partial, 6 missing
Why. Decision 6
Before. Curated data at version 1.2
After. proposals, key figures, glossary and sources brought into line: fee subsidy, Work Pass deferred, apprentice pay, relief notice on the Joint Operator Scheme, a new 'no compulsory designated suppliers' proposal, 776 coffeeshops, the HDP's two baselines, re-read model figures, four new glossary terms and 54 new sources
Why. Decisions 1–3, 6 and 7; build_data.py --strict passes
Before. Median AMR ≈$1,200; $1,450 the median rent non-subsidised stallholders paid
After. Median AMR $1,250 (13 Nov 2024); system median rent about $1,250, stable for a decade; $1,450 the 2025 median at centres comparable with the operator-run ones
Why. W1 #5, W3 #24, W5 #11, W5 #12
Before. 2.8%: HDP graduates still operating a stall (16 of 566); 'a handful operating'
After. Of 566 who finished classroom training, 16 run a stall; 29 started one, 16 still operate
Why. W1 #26
Before. 'No repair fund and no ducting programme anywhere'; repairs and exhausts 'the landlord's'; HPG 'closed'
After. The State repairs and upgrades centres through R&R every 6–8 years and HUP 2.0 (Mar 2025, up to $1bn, ventilation included); the stall's duct is the stallholder's legal duty; the gap is stall-level equipment and duct upkeep while the HPG is between phases; new §1.3 row
Why. W1 #40, W2 #7
Before. CDC tranches $300 and $500; NEA does not require budget meals; personal operation with 'flexibility'
After. Adds the $300 tranche for Jan 2027; the SEHC value-meal tender condition; the Joint Operator Scheme and nominee route; PHC dishwashing co-funding
Why. W2 #13, W5 #24, W5 #37, W2 #8
Before. Missed helpers; FMAS awards with no prize money; CDC $1.06bn (May 2025); KF Seetoh 'exposed' the clauses
After. Adds HawkerBoss, Fei Siong's programmes, ITE, Feed the City, MacPherson's $3 vouchers, Temasek Polytechnic's 2015 book, Maybank's Hawkers' Omakase, Deliveroo's exit; S$500 prizes; $2.1bn (MCCY, Feb 2026); 'raised ... disputed'; Belanja status unconfirmed
Why. W3 #28, #30–#34, #39–#41; W2 #14; W6 #31
Before. Town Council collects the conservancy and table-cleaning fees; the fee income given up is its own
After. It collects the conservancy charge; the table-cleaning fee follows the cleaning contract; re-tendering needs its agreement; the subsidy is the Government's
Why. W1 #4, W5 #10
Before. 'What no one does yet': 'no one pays ... goodwill', 'no one offers ... income top-up', recipes
After. 'What no scheme does yet': no scheme pays goodwill, the few private sales are one-off deals (Sin Kee, $42,800 each, 2016); no scheme tops up the hawker's own income (HawkerBoss the nearest); no standing archive good enough to cook from; ComCare and private medical-leave cover named
Why. W3 #42, #44, #32; W2 #33
Before. Not new: the ballot (Leong, flat rent); buying together; stipends
After. Adds Hazel Poa, Dr Koh's four objections, the amended motion with PSP support; a foreign hand at the stall; the WP adjournment motion of 3 Aug 2021; the PSP's 2024 call for targeted discounts; HSS and Stewards amounts spelt out
Why. W3 #2, #3, #6, #13, #21, #22; W2 #6; W5 #55
Before. Goodwill on the HSS rail 'within existing scheme parameters'; 'coffee money' changes hands
After. NEA must amend the HSS terms to allow a disclosed, capped goodwill payment for recipes, training and name, never the location; Kay Lee ($4M, 2014) and Sin Kee as precedents, with the food-drift lesson
Why. W5 #18, W5 #56
Before. 'as a platform now shares the CPF cost for a rider'; 'The rider got CPF'
After. For riders born in or after 1995 and the roughly 27% of older riders who opted in; every rider has work-injury cover
Why. W2 #43
Before. Stretch match ≈$8.0M; base shortfall Years 4–10 $8.8M, stretch $26.8M ($31.7M without HDB); base short from Year 7; ~$1M a year of gifts set aside; corpus Y20 118% / 160% / 86%; HDB lever ≈$1.8M, under-55 part ≈$0.9M
After. ≈$7.9M; $10.9M, $28.6M ($33.2M); from Year 6; ~$0.7M; 106% / 133% / 86%; ≈$1.7M, ≈$0.84M; the year-by-year table re-run
Why. numbers_M.md
Before. Tendered NEA stall at $1,450, 'what sitting tenants actually pay'; quoted rents $1,200 AMR, $1,450, $1,800
After. Tendered stall at the $1,250 median across NEA's centres (also the median AMR); $1,450 kept as the 2025 median at centres comparable with operator-run ones; quoted rents $1,250, $1,450, $1,800
Why. W4 #11, W1 #5, W1 #32, W6 #26: $1,450 is a comparable-centre median (MSE, 8 Sep 2026); the system median is about $1,250
Before. Delivery commission 22% (Monte Carlo 15–29%), from a vendor blog
After. 30%, GrabFood's published fee before GST, varied 25–35%; Deliveroo's exit noted
Why. W4 #30 (high): Grab publishes 30% plus GST
Before. Buying Circles' delivery at 15% instead of 22%
After. Counted at a 25% group rate against the 30% list rate; 15% kept as the circles' target, shown as a sensitivity
Why. W4 #31: 15% is unsupported and half the published rate; test 20–25% as the central case
Before. Table-cleaning fee reset booked as fee income the collector gives up; operators bear 45% at SEHCs
After. A $200-a-month fee subsidy the Government pays to the cleaning contractor at pilot-centre stalls, all public money, an upper bound tested by re-tendering; landlord take unchanged
Why. Decision 1; W1 #30, W1 #31, W5 #9; MSE 6 Jul 2021
Before. SEHC: 60% of fees landlord income; operator keeps 45% of rent and fees
After. Table cleaning a pass-through (30% landlord share on S&CC only); operator share 10%, derived from MSE's ~$100,000-a-centre surplus at an assumed 48 stalls
Why. W6 #20, #21; W1 #30: operators may not profit from pass-through charges (NEA, 21 Jul 2026)
Before. Protection Floor cuts MediSave arrears from a quarter to one in twenty, adding about $2.0M a year of Workfare
After. Arrears assumed one in twenty with or without the floor, since licence renewal already enforces MediSave; no Workfare uplift
Why. W2 #19 (high): SFA and CPF enforce MediSave at licence renewal
Before. 805 HDB coffeeshops (405 rented, 400 sold); 3,000 HDB-rented coffeeshop stalls; 14,400 stalls in all; HDB lever up to $1.8M
After. 776 (374 rented, 402 private, July 2024); 2,800 stalls; 14,200 in all; HDB lever up to $1.68M; tender receipts about $215M, an upper estimate
Why. W1 #36, W6 #24, W4 #20
Before. Relief cook paid a fixed $190 a day; Corps member $12,500 a year
After. PWM Cook (QS) rate for a relief day of at most 12 hours, the 4 beyond the normal 8 at 1.5 times, with 17% employer CPF: $212 (2026), $224 (Year 1), $236 (Years 2–3); member $14,753 then $15,539
Why. W4 #49: index to the PWM cook rate (MOM, 16 Mar 2026)
Before. Wage shock $280 a month without employer CPF; one utilities row, −10% for the Q4 2026 cut
After. Every wage +12.6% on the PWM path, CPF included ($368 at a tendered stall); two energy rows: Q3 2026 +13% (half passed on) and the Q4 cut −10% on Q3; Q1 2027 rise expected
Why. W4 #6, #25, #26, #27, #29
Before. COVID replay: three months' rent waiver in 2020 and one in 2021
After. Eight months' rent waiver and six months of table-cleaning fee subsidies from April 2020; with SIRS the protected income now carries the 2020 hand-over through
Why. W2 #12: MSE, 3 Aug 2021
Before. Supply-chain split 'an assumption, labelled as such'
After. Marked 'an assumption, not data' on Figure D.1 and in the text: 65% landed value, 35% Singapore-side, the middle of a 25–45% range
Why. W6 #8, analysis §5.3
Before. The circles move about $12.0M a year 'from suppliers' margins' to hawkers
After. Circle members pay suppliers about $11.8M a year less: a lower cost of serving many small buyers, shared between supplier and cook
Why. Decision 6; W6 #28; margins tracked import prices 2019–25
Before. No section on middlemen
After. New D.5.3, 'The Middlemen: Who Takes a Cut Between the Farm and the Plate', each number labelled official, derived or assumption; asks NEA to bar compulsory designated suppliers at operator-run centres. No id moved
Why. Decision 6; W6 analysis and draft
Before. Seventeen indicators in the Plate Ledger
After. Eighteen: adds the margin spread (staple prices at wholesale and retail against the food import price index, quarterly)
Why. Decision 6
Before. SingStat 2021: 7,050 establishments, $4.1bn; 1.49M households 'HES 2023'; July 2026 rent support without dates or purpose
After. Softened to 'roughly $3–4 billion', unconfirmed; 1.49M from SingStat 2025; support dated Sep 2026–Feb 2027, to offset energy costs, left out of the baseline
Why. W4 #39, #40, #41, #42
Before. Coffeeshop rents from asking-rent listings (median about $3,400)
After. $3,400 labelled an assumption; a commercial guide gives $4,000–10,000 (2026), so the base is conservative
Why. W1 #34, #35
Before. MediSave and income-tax relief at the 50+ and 60+ rates for every stall
After. By the stallholder's age band (MediSave 8–10.5%; Earned Income Relief $1,000, $6,000 or $8,000); archetypes unchanged at 60
Why. Needed for the prologue's cast (decision 5); CPF and IRAS rates
Before. No model support for the /day/ page
After. day_view, centre_view and landing_top_up, with John (single-handed) and Ah Kow (400 plates by two) as cases, ported to the simulator and parity-tested; results.json 'day' for three states
Why. Decision 5; sim_design §7.3
Before. Simulator labelled 'Technical Paper', 'Main Paper §2', T1.1.7, T1.2.2, T1.2.4, 'Framework §7.6'
After. Follow the Plate (Volume D), the Plan Part II, C.7, D.3, D.5, the Budget Book B.9
Why. 1.1 labels left after 1.2
Before. No section on how a stall's costs moved over time
After. New D.5.4, The Plate Over Time: How a Stall's Costs Changed, 2013–2026, with Figures D.6 (costs indexed to 2019) and D.7 (cost shares 2012–13, 2022, 2026)
Why. Tests the claim that government layers drove up hawker costs: partly true in form, not supported as a driver of prices; the narrow complaints (the operator-run floor-cost gap, delivery) hold
Before. Figure D.6, incidence of stall-attached support
After. Figure D.8 (the two figures of D.5.4 are D.6 and D.7, in reading order)
Why. Figures numbered in reading order; no section id changes
Before. Eighteen tracker indicators (3 published, 9 in part, 6 not at all)
After. Nineteen: adds floor costs by type of centre (rent, conservancy, table cleaning and dishwashing, before GST), in part (MSE's 2025 medians)
Why. D.5.4: the $2,750 against $2,170 floor-cost gap at operator-run centres (MSE, 8 Sep 2026) should be tracked, not argued about
Before. NEA re-tenders the pilot centres' cleaning contracts; who holds the contract at a Town Council centre to be confirmed; the Town Council collects the table-cleaning fee
After. The contract's holder re-tenders, convened by NEA: at HDB hawker centres (Chinatown Complex and Tekka among them) NEA's notices say the table-cleaning service 'is managed by the respective Hawkers Association'; at Geylang Serai (NEA) Year 0 confirms the holder
Why. NEA tender notices, Dec 2022 and Aug 2024, Note 1(d), confirmed on 29 Oct 2026; the note covers HDB hawker centres, not NEA's own; Aug 2026 notice names only 'the table-cleaning contractor'. The subsidy, its amount and its payer are unchanged
Before. The fee follows the contractor's price; the cleaners' PWM rose from $1,312 to $2,080
After. Adds: at four NEA-managed centres the fee was the same before GST in Dec 2022 and Aug 2024 ($402 at Amoy Street, $360 at Chomp Chomp) while the cleaner's PWM rose by a third ($1,312 to $1,740); a re-tender may find no saving
Why. Derived from NEA's tender notices (fees printed with 7% and 9% GST) and MOM
Before. There is no Singapore estimate of how diners respond to hawker prices
After. One early, indicative estimate: Poh (2026), a stated-preference survey of 90 diners, not random; elasticities 2.81 (18–24), 1.01 (45–60), 3.92 (60+, nine people), 2.09 on low meal budgets against 1.1–1.2; the model's −0.5 is unchanged
Why. Poh, Decoding the Consumers of Hawker Food (Scholarly Review Journal, Fall 2026); no model parameter changes
Before. IPS Makan Index 2.0 … (2023)
After. (prices of late 2022, published 2023)
Why. Fieldwork Sep–Nov 2022, revisit Jan–Feb 2023; IPS Exchange Series No. 24, March 2023
Before. D's sources: no group for the plate over time
After. Adds 'The plate over time (D.5.4)': SingStat, EMA, SP Group, NEA (incl. the Dec 2022 and Aug 2024 tender notices), MSE, MOM, IPS (Working Paper No. 29; Exchange Series No. 24) and Poh (2026); the volume's sources note names what D.5.4 could not confirm
Why. Every new fact cited; back-casts and secondary figures marked
Before. Coffeeshops and food courts against hawker centres compared only by price rises (15.1% against 15.8%, 2019–23)
After. Adds the level: IPS's 2022 averages, food courts 18–36% dearer than hawker centres for the same dish, kopitiams close on dishes and about 14% dearer on drinks; not all margin (air-conditioning, mall and turnover rents; menu prices, not adjusted for portion); SingStat's yearly rates given
Why. IPS Exchange Series No. 24, Table 3.1; SingStat Newsletter, Issue 1, 2024 (source PDFs added under sources/)
Before. Private coffeeshop landlords: no figures
After. Adds KF Seetoh's view, attributed and marked as press report: coffeeshops sold for over $40M in June 2022 (Tampines $41.68M), stall rents of about $12,000 a month, at least 10 of 14 Yishun stalls gone; his call to end highest-bid tendering
Why. The Online Citizen, 16 Mar 2023 (copy hosted by IPS)
Before. MTI 2015 cited for its pass-through coefficients only
After. Adds its stall-level panel (2012–14): about two in three stalls that raised prices had no rent change; subsidised and non-subsidised stalls priced alike; higher pass-through in the CBD; one more similar stall, prices about 8.4% lower; publishing prices was among its measures
Why. MTI, Economic Survey 1Q2015, checked against its text; cited at its current URL
Before. NEA's rent flat at about $1,250 from 2015
After. About $1,250 for a non-subsidised stall in 2012–13 (MTI's average) and 2015–23 (MSE's median); the Makan Index comparison marked indicative (the 2.0 report never refers to 2017's), with IPS's own revisit of 263 stalls
Why. MTI 1Q2015; IPS Exchange Series No. 24, Table 3.56
Before. The relief cook paid the PWM cook rate for the stall's whole 13-hour day, at straight time, with employer CPF (about $197 in 2026)
After. A relief day of at most 12 hours: 8 normal hours at the PWM Cook (QS) rate and 4 at 1.5 times, with 17% employer CPF, about $212 in 2026, $224 from July 2027 and $236 from July 2028; on a 13-hour stall day the stall's own assistant does the first hour's prep, and a stall with no assistant opens an hour later
Why. The corps member is an employee: the Employment Act (Part 4) caps the day at 12 hours, sets normal hours at 8 a day for a week of more than five days and requires overtime at 1.5 times (MOM). Review B1
Before. Corps member $13,699 (Year 1) and $14,429 (Years 2–3); Corps lines $340,281, $653,873, $711,589; the Corps' share of the rise about $0.19M
After. Corps member $14,753 and $15,539; Corps lines $360,307, $694,943, $757,099; the Corps' share about $0.31M of a $3,872,802 rise; the Relief Corps' leverage 2.17 (was 2.45); national package $78.7M a year, leverage 1.06
Why. The model re-run with the lawful relief day (plate_model.py, funding_model.py; both parity tests pass)
Before. From Year 4 … no year runs short
After. At the central assumptions no year runs short; in about one future in eight some year does
Why. In the Monte Carlo 12.7% of futures have a short year. Review S12
Before. Grace Fu kept work permit holders out of hawker centres "to preserve the local identity and heritage of our hawker culture"
After. Asked why work permit holders may work in coffeeshops but not in hawker centres, the Minister left the rules as they are: "Hawker centres are a unique part of Singapore's cultural heritage and it is important to preserve the local identity" (MSE, 10 Sep 2026); MOM's rule "To preserve the local identity of our hawker culture and heritage" (25 Feb 2026); both sources added
Why. Each quote now matches its source, and each claim says only what the reply says. Review S2
Before. the table-cleaning fee follows whichever cleaning contract it or NEA lets
After. the fee follows the cleaning contract, which at HDB centres the hawkers' association manages; re-tendering needs the association to act and the Town Council to agree
Why. NEA tender notices, Dec 2022 and Aug 2024; matches the rest of the library. Review S4
Before. Rent bands … leverage 0.90
After. leverage 0.83
Why. The 1.3 model gives 0.83 (D.7). Review S5
Before. the same dish cost a quarter to a third more at a food court; kopitiams about 14% dearer on drinks
After. from 4% to 41% more across IPS's 18 items, a median of about a quarter; the seven dishes named as a selection; kopitiams 11–14% dearer on drinks
Why. IPS Exchange Series No. 24, Table 3.1. Review S6
Before. a stipend of at most about $5,500
After. about $5,000 in all
Why. 3 × $1,500 + $500 = $5,000, as the same section says. Review S7
Before. Source headings print source ids: "The middlemen (B.5.3)", "The plate over time (B.5.4)", "(B.10.6)"
After. D.5.3, D.5.4 and "the Budget Book, B.9.6": compose.py now rewrites the ids in a volume's own source lists
Why. Readers never see source ids (AGENTS.md). Review S8
Before. Delivery row: "$0.11 on average; $1.35 on a delivered plate", overflowing its cell
After. "$0.11 average / $1.35 delivered" on two lines; the table may break across pages
Why. The figure $1.35 could not be read. Review S9
Before. Suppliers take the most; 45¢ of each dollar to ingredient suppliers; Suppliers take the largest share; column "Supplier take"
After. Ingredients cost the most; of each dollar on ingredients; Ingredients are the largest share; column "Ingredient share"
Why. Most of the 45¢ is farm or import value, not a supplier's take (D.5.3). Review S10
Before. NEA, the landlord at 107 centres … kept the conservancy charge at $140 … the State as landlord held still; Figure D.6: a stall at an NEA or Town Council centre
After. NEA's $140 → $155 → $175 is its charge at the 27 centres it manages; Town Councils set their own; the State held its rent still and raised its conservancy charge by a quarter, citing cleaning costs; Figure D.6 retitled
Why. NEA, 3 Jun 2023; NEA tender notice, Aug 2026, note 1(e). Review S11
Before. Grounded: every claim checked against its sources and corrected where wrong
After. Grounded: the library's claims checked against sources (314 checked; corrections made where the evidence differed)
Why. Each volume names claims that could not be confirmed. Review S14
Before. no one keeps a list; which no one counts today; Fills a gap no one fills; No one keeps a hawker's recipes; No one finances or guarantees them; No one publishes those
After. Government keeps no list; which Government does not count and we found no one else who does; Fills a gap (none found); No scheme keeps …; No scheme finances …; No one publishes them regularly; only one-off medians and cost shares exist (MTI 2015; NEA 2022; MSE 8 Sep 2026)
Why. Say what the evidence supports. Review S15
Before. about $399 a month more profit
After. about $407 a month more profit ($4,791 a year in income and savings)
Why. $399 was the after-tax household gain. Review nit 1
Before. 44¢ on the opener and in Figure D.2, 45¢ elsewhere
After. 45¢ throughout, rounded half up from 44.5%
Why. Review nit 2
Before. break-even about 10 hours; 2022 and 2023 both 'the steepest since 2008'; kopi-O about 13% more; most rises 10 cents or less; seven price rises of 10–30 cents a kilo; SP Group expects a Q1 2027 rise
After. about 11 hours (10.8); only 2023's rise the steepest since 2008; about 12%; for most food items the average rise was no more than 10 cents; the cartel described as CCCS does, coordinating the amount and timing of price rises; one analyst (Rystad Energy, as reported) expects a single-digit rise in Q1 2027, SP Group has not said
Why. SingStat; IPS Exchange No. 24; CCCS, 12 Sep 2018; The Star, 30 Sep 2026. Review nits 3, 4, 5, 7, 20
Before. Deliveroo left Singapore on 4 March 2026
After. Deliveroo went offline in Singapore after 4 March 2026
Why. It stayed live until 4 March. Review nit 8
Before. HawkerBoss … in Kopitiam and Foodfare outlets; Kopitiam and Foodfare as partners
After. in Kopitiam outlets; Kopitiam as a partner
Why. The 10 May 2022 release names Kopitiam outlets only. Review nit 6
Before. apprentice stipends; $31.7M of public money (anchor, match, tax forgone and operating match); about $18.7 million
After. apprentice wages; NHB's half of the Year 0 research grants ($321K) listed; about $18.86 million
Why. Review nits 9, 11, 15
Before. the one in four older riders; between HPG phases, nothing funds it; I agree with the PSP; one joint operator; at any sane price; nearly half of Singapore's hawkers; Chinatown's start timed after its works
After. more than one in four; the HPG co-funds new equipment, never repairs; I agree that stalls should be allocated by ballot (and with the proposal); a joint operator; clause removed; about two in five; the ledgers assume the works are done before Year 1 begins in July 2027
Why. Review nits 13 (Figure D.6), 14, 16, 17, 18, 19, 21 and the unsourced §2.9 clause
Before. Figure D.6: an event label ran across the 2023 rule; the conservancy line hid under NEA's rent
After. Labels placed in rows that cross no later rule; the rent drawn wider beneath the conservancy line
Why. Review nit 13
Before. (no independence statement in the PDFs)
After. An Independence row on every Document Control page, and on the change log's: the disclaimer of website/BRIEF.md §7.6
Why. AGENTS.md requires the disclaimer on every public page; the PDFs lacked it. The Document Control page is set tighter so it still fits one page; the Executive Summary stays at 8 pages.
Before. only 4% of the 7,000 cooked-food stalls pay rent above the assessed market rent of $1,250
After. only 4% pay above their own stall's assessed market rent, whose median is about $1,250 (Dr Koh's correction by written statement, 13 Nov 2024)
Why. Dr Koh corrected the figure the same day (Official Report, SG Parliament debate 27997); each stall is compared with its own AMR.
Before. Quality tenders, the foreign hand and the cheap plate credited mainly to the PSP and Louis Chua
After. Credited even-handedly: Perera's 2021 discount cards and Chua's CHAS discounts repeated at COS 2026; Tan Yia Swam (2023), Jessica Tan and Foo Cexiang (2026) and Jamus Lim (6 Oct 2026) on a foreign hand, answered at least six times; a new 'Quality in the tender' item (Perera, Chua, Liang Eng Hwa, Mariam Jaafar, Jamus Lim; HDB's 40 of 65 tenders to other than the highest bidder)
Why. The Hansard review (29 October 2026) found cross-party proposals uncredited. Louis Chua's credits of ba2474b are kept and not repeated.
Before. (no Government answer to the academy proposals)
After. Dr Koh's reply, "we already have a similar academy in place" (the Asian Culinary Institute); the apprenticeship takes HDP entrants and adds the master's side
Why. Even-handed attribution: the Government's answer belongs beside the proposal.
Before. Melvin Yong asked whether a ballot at a fixed rental rate had been considered
After. a ballot "at a fixed rental rate targeted at cost recovery"
Why. The Official Report of 14 October 2024 (SG Parliament 27520); the phrase distinguishes it from a turnover rent. Same fix in proposals.json, whose note also wrongly called Leong's a flat rent.
Before. has twice declined rent control (2022, 2024); the rent-control fight ministers have declined twice
After. declined repeatedly: a hawker-centre cap in 2018 and 2024, coffeeshop rents in 2022, all commercial rents on 6 October 2026 (Low Yen Ling: "imposing rental controls is not straightforward and could have far-reaching consequences")
Why. The record shows more refusals than two, the latest on 6 October 2026 (SG Parliament 37232).
Before. The 2024 debate cited from the party pages, and the Official Report for the motion and resolution only
After. The Official Report (SG Parliament mirror, debates 27944–27948, and the correction 27997) cited beside the party pages, marked as an unofficial reproduction; the 2012, 2026 and COS references added
Why. Primary record first, party copies second.
Before. 2013–19 prices rose about 12% (a back-cast); Figure D.6's prices dashed to 2018; NEA rent line to 2023; 'no system-wide median after 2023'
After. about 12.5%: official rates for 2013 (1.9%) and 2014 (3.0%), 2015–19 about 1.8% a year, checked against the FSI deflator; rent line to 2024; NEA restated the ~$1,250 median on 4 Nov 2024
Why. MTI's reply of 13 February 2015 gives the official rates; NEA's release of 4 November 2024 restates the median (R1, R3, R4).
Before. The 2013–19 residual sits where the data cannot see
After. adds the timing: F&B rents outside hawker centres +7.4% a year in 2010–14, −2.6% in 2015–19; prices did not come down when costs did (my reading, not a measurement); shares: utilities about 7% in 2022–24, rent's share 14% to under 10%; the subsidised share ran down from about half to 30% and the ownership leases returned to rent
Why. T2 revisions R2, R5, R6 (MTI, 1 Aug 2022; MSE, 4 Aug 2026; NEA).
Before. (no revenue side)
After. 'The other half of the purse': FSI real sales flat 2012–25, down 1.5–5.3% in Feb–May 2026; takings +31% against pay +46% (2015–25); Figure D.6b plots price, plates and pay
Why. T2 revision R9: the squeeze is in plates and relative income, not the price-cost margin. New figure D.6b; no other figure renumbered.
Before. tray return 'the rule since 2021'; operator-run centres without the app discount; HES fall 'about 13%'
After. tray return enforced from 1 Sep 2021; stallholders' $1,000 a month app discount (as reported) and value meals at 5–20% of meals; SingStat's item prices 2014–24; 'about 13–14%'; the closing adds the plates; the Plate Ledger box adds a plates line and asks NEA to publish its yearly stall-cost survey
Why. T2 revisions R7, R8, R10–R13.
Before. Assistants' wages explain 2–5 points (no word on the PWM's reach)
After. Individually run stalls are exempt from the Food Services PWM, which is enforced through work-pass hiring; it sets the going rate, not a floor
Why. T2 revision R15 (MOM and TCF, 16 Mar 2026).
Before. Two tests of the middle; '402 coffeeshops HDB has sold'; delivery fees 'expected to rise'; utilities 'about a point'
After. A third test on operators' books (rent 17.0% to 10.8% of revenue, surplus 3.2% to 2.3%, 2019–24); HDB's 374 rented coffeeshops, 45% to chains, and the January 2026 budget-meal change; CCCS saw no systematic shift in commissions by April 2026; utilities held at about 7%
Why. T2 revisions R14, R16–R18 (MTI, 7 Apr 2026; MND, May 2024).
Before. Hawker food prices: keep; add a split by type of landlord. Cost shares: NEA survey, 2022 (one-off)
After. Print real sales beside the price index; cost shares from NEA's yearly survey (quoted for 2022–24, unpublished): publish it
Why. The plates line and the yearly survey (R13), without adding an indicator.
Before. (no layers timeline)
After. The layers timeline, 1966–2026: 36 layers between the State and the wok, with what each did to costs or takings
Why. It does not fit in D.5.4 within Follow the Plate's length; D.5.4 points to it.
Before. utilities share and pre-2019 prices unconfirmed
After. Utilities confirmed for 2022–24; 2013–14 prices official, 2015–18 assumed; the FSI is a proxy for a stall's plates; item prices and the app discount secondary; 23 new sources listed in D; 34 new entries in data/curated/sources.json
Why. Each volume's sources note says what could not be confirmed.
Before. Three pilot centres, Years 0–3: $18,863,142; gaps $12.76M; ladder $12.21M; corpus 133% at Year 20; keeps real value in 78% of futures
After. $19,581,112 (+$717,970; +$4,590,772 on 1.2); gaps $13.48M; ladder $12.90M; corpus 129%; 76% of futures; from Year 4 $4.13M a year
Why. Re-run of funding_model.py with the new Fellowship terms and the Regulars' List costing
Before. Family Name Royalty starts when the mortgage is retired, with no end date
After. Runs for ten years at most, with a buy-out at any time at a published multiple; the cap never binds inside the stress test's ten-year horizon, so no E figure changes
Why. Uncle James, 29 Oct 2026: the royalty is capped at ten years, no longer perpetual
Before. The National Hawker Endowment; 'The Heritage Watchlist' as a parameter label
After. The Hawker Endowment; 'The Regulars' List'
Why. Uncle James, 29 Oct 2026: printed names (the slug national-hawker-endowment is kept)
Before. year one of HSS 2.0's three-year rent glide
After. year one of the extended Hawkers' Succession Scheme's three-year rent glide
Why. Uncle James, 29 Oct 2026: 'extending the Hawkers' Succession Scheme' replaces 'HSS 2.0'
Before. (none) the Watchlist had no tiers
After. New D.4.2, 'The Regulars' List: What a Tier Means in Plates': Bronze (the listing) always 20 named regulars; Silver 50, Gold 100, Platinum 250, or 8%, 15%, 40% of weekly plates, whichever first; a 35% healthy base (ASSUMPTION); table for every archetype and for John and Ah Kow
Why. Uncle James, 29 Oct 2026 (08:45): tiers by endorsement count, and a simulation of the share of plates from regulars a healthy stall needs
Before. plate_model.py had no regulars measure
After. regulars_tiers() and regulars_suite() in plate_model.py (results.json 'regulars'), ported line for line to sim_src/model.js with parity cases (edited stalls at 0.5–3 visits a week)
Why. Model and browser engine stay in parity (AGENTS.md); NRA 2013, Toast and Resy 2026 and NEA's 2018 patron survey cited
Before. FMAS invited to name a non-voting observer at Council meetings
After. FMAS names the holder of a tenth Council seat, with a full vote
Why. Uncle James, 29 Oct 2026: FMAS gets one voting seat on the Hawker Council, not an observer.
Before. A recognised voice on tenancy and operator-contract terms · No bargaining body
After. A recognised voice: FMAS and the centre hawkers' associations, recognised as the trade's bargaining body, not a new one · No bargaining body with a mandate
Why. Uncle James, 29 Oct 2026: the bargaining body means FMAS and the centre associations.
Before. an ombudsman reporting to MSF
After. an independent ombudsman, appointed by the board and reporting publicly each year
Why. Uncle James, 29 Oct 2026.
Before. Fellowship: $3,500 a month for 12 months ($42,000), CPF $5,250, $77,250 a Fellow over one year
After. A teaching year at $2,500 a month ($30,000), then a 24-month stewardship at $1,000 a month ($24,000) for light duties (tastings, locum tests, two visits a month to the apprentice's stall, one public session a year); completion grant and health cover unchanged; CPF 12.5% on the whole stipend ($6,750); $90,750 a Fellow over three years, each cohort's three years budgeted in the year it starts; intensive track compresses the teaching year then the same stewardship; Emeritus pathway follows the stewardship
Why. Uncle James, 29 Oct 2026: a Fellow should not fall off a cliff when the teaching year ends; less each month, for longer.
Before. The Heritage Watchlist: four qualifying tests, a verification visit, a yearly four-question check-in with $30 + $10 visit stipends, triage on two signals (65+, no successor, tenancy ending, last stalls), decal 'On the Heritage Watchlist'
After. The Regulars' List: a compliment from diners ('the regulars of this centre have asked that this dish be kept'); listed at Bronze on 20 endorsements by weekly regulars (one each, verified by a mobile number, never bought) or an association nomination; Silver, Gold and Platinum at higher counts or plate shares, whichever first; tiers only go up; leaves only by the hawker's choice or closing; no re-test; the same small things for all (Registry page, decal with tier and year, price board and order book, booklet and map, thank-you note); no stipends or decal fees; Fellowships opt-in, offered when the hawker wants to teach and the Registry's open test finds the dish fragile, never gated by tier; the at-risk count kept apart, from two voluntary signals (no successor in view, intention to stop within three years) and public facts, published only as national and per-centre totals
Why. Uncle James, 29 Oct 2026 (08:30 and 08:45): simplify the working assumptions, make the list a compliment that does not hurt a stall that stays on it, and tier it by endorsements.
Before. Hawkers Succession Scheme 2.0
After. The Hawkers Succession Scheme, Extended (heading); 'extending the Hawkers Succession Scheme' elsewhere; section id and proposal id hss-2-0 kept
Why. Uncle James, 29 Oct 2026.
Before. Family Name Royalty: 2.5% of gross, in perpetuity ('forever', 'a permanent stake')
After. 2.5% of gross for ten years at most from its first month (about $55,000 for a $220,000 stall), with a buy-out at any time at a multiple the Seed Bank publishes in advance; why ten years said plainly
Why. Uncle James, 29 Oct 2026. The stress test is unchanged: the cap never binds within its ten-year horizon (numbers_A3 §12).
Before. Nineteen indicators in the Plate Tracker (data/tracker.json)
After. Twenty: t20 'Stalls at risk, and the Regulars' List' (national and per-centre at-risk totals, never naming a stall; listed stalls by tier), status missing, closed by the Seed Bank's yearly count from the end of Year 1 (§3.6)
Why. Uncle James, 29 Oct 2026 (08:45): carry the tiers and the at-risk count into the tracker. B.9's TRACK in build_addB.py needs the same row.
Before. In this volume: … D.4 What the hawker earns per hour …
After. … D.4 What the hawker earns per hour, and what a Regulars' List tier means in plates (D.4.2) …; a 'Regular' term; the visit rate and the 35% healthy share named as assumptions in the sources note
Why. The new D.4.2 (worker A) is listed where the volume lists its sections.
Before. Nineteen Plate Tracker indicators: Singapore publishes 3 regularly, 10 in part and 6 not at all
After. Twenty: a new row, 'Stalls at risk, and the Regulars' List' (the voluntary at-risk count as national and per-centre totals, never naming a stall, and the listed stalls by tier), missing today; 3 published, 10 in part, 7 not at all
Why. The at-risk count of §3.6 is published only as an aggregate; the Tracker carries it so that the count and the table stay in step
1.2 28 October 2026 10 changes · shared version
A restructure for readability, with no change to any proposal, figure or claim. The three papers become a library of standalone documents sharing one version: the Executive Summary; the Plan, the case and every proposal; and six companion volumes – Who Already Helps (A), the Budget Book (B), Hawkers and Social Protection (C), Follow the Plate (D), Succession Economics (E) and Method, Levers and Data (F) – each opening with who should read it, its bottom line in five points, how it connects and the terms it uses. The Main Paper and the Technical Paper end; every section moves, with a new id, and every cross-reference names the document it points to. Under the versioning rule Uncle James set on 28 October 2026, a restructure by the initiator alone is a minor version; a new edition, with contributors or reviewers beyond him, is a major one.
Approved by Uncle James · tag v1.2 · release commit 298c540
Before. One document in three papers: the Executive Summary (8 pp), the Main Paper (86 pp) and the Technical Paper, Under the Hood (73 pp), with the Change Log beside them.
After. One document as a library of nine PDFs: 1 Executive Summary; 2 The Plan; A Who Already Helps; B The Budget Book; C Hawkers and Social Protection; D Follow the Plate; E Succession Economics; F Method, Levers and Data; 0 the Change Log. The Main Paper and Technical Paper PDFs are removed; the repository keeps one copy of each document.
Why. Uncle James (28 Oct 2026): "separate each appendix chunk into standalone PDFs, so that the download page by itself makes sense, with each major appendix block being a document with standalone reading viability. I'm keen to significantly uplift readability regardless of where they start." Moving the detail into one appendix would only have made the appendix too large.
Before. §1.3.1 Beyond Government: Who Else Already Helps and §1.3.2 What Only This Plan Does, in the Main Paper.
After. Who Already Helps: A.1 Standing on What Works (§1.3's rails, reprinted so the volume reads on its own), A.2 (§1.3.1) and A.3 (§1.3.2).
Why. A partner or charity can read the whole review in one short document.
Before. The ledgers, the endowment, the ask and the downside in Main Paper Part VII; Follow the Gift in Technical Paper T1.4, with the heritage-stall method in T1.4.6; the gates and funding levers in T2.7–T2.8.
After. The Budget Book: B.1 Reading the Ledgers (§7.1, with each centre's budget and the add-ons), B.2–B.5 the ledgers for Years 0–3 (§7.2–§7.5), B.6 the endowment (§7.6, with B.6.1 and B.6.2), B.7 the ask itemised in full (§7.7), B.8 downside mode and sunset in full (§7.8), B.9 Follow the Gift (T1.4; the heritage-stall method is B.9.6), B.10 and B.11 the levers (T2.7, T2.8).
Why. A funder can read every line of the money, and the model that tests it, in one document.
Before. Social protection compared in Technical Paper T1.1 (T1.1.1–T1.1.7); the Protection Floor's levers in T2.6, with the retirement match costed in T2.6.1.
After. Hawkers and Social Protection: C.1–C.6 (T1.1.1–T1.1.6), C.7 Whoever the Landlord (T1.1.7), C.8 the levers on the Protection Floor (T2.6) and C.8.1 the retirement match (T2.6.1). §2.11, the Floor itself, stays in the Plan.
Why. Labour and CPF policy readers get the comparison and its levers together.
Before. Follow the Plate in Technical Paper T1.2 (T1.2.1–T1.2.8); the levers on rent bands, vouchers, the Relief Corps and the master's side in T2.1–T2.4, and on Buying Circles in T2.9.
After. Follow the Plate: D.1–D.7 (T1.2.1–T1.2.7: stall economics, the rent band evidence in D.4.1, scoring the proposals in D.7), D.8 the Plate Tracker (T1.2.8), D.9–D.12 (T2.1–T2.4) and D.13 Buying Circles (T2.9). Its figures are numbered D.1–D.6.
Why. NEA, economists and hawkers' associations get the stall economics and the levers that rest on them together.
Before. The succession stress test in Technical Paper T1.3; the Succession Mortgage's levers in T2.5, with the worked examples across all three tenures in T2.5.1.
After. Succession Economics: E.1 Can the Business Pay for Its Own Succession? (T1.3, sub-sections E.1.1–E.1.5), E.2 the levers on the Succession Mortgage and the Family Name Royalty (T2.5), E.2.1 Across All Three Tenures (T2.5.1). Its figure is E.1.
Why. A hawker family, a bank or a lawyer can read the hand-over's economics on their own.
Before. How to read the levers opened Technical Paper T2; T3 held the limits and assumptions, the model parameters, the data and downloads, and the references of the three papers.
After. Method, Levers and Data: F.1 How to Read the Levers, with a table of where each lever is now printed; F.2 Limits and Assumptions (T3.1); F.3 Model Parameters (T3.2); F.4 Data and Downloads (T3.3); F.5 References for the whole library, consolidated (T3.4). The Plan and each volume keep a short sources note; Hawkers and Social Protection and Follow the Plate keep their own source lists.
Why. A researcher or evaluator finds the method and every reference in one place.
Before. —
After. Each companion volume opens with a one-page opener in Uncle James's voice, drawn from its body: 'Read this if you…', 'The bottom line' in five points, 'How this connects' (where it comes from, what to read with it, what is in it) and the terms it uses; and closes with a short sources section and its own Document Control page. The text is kept in src/volumes/. No proposal, figure or claim is added.
Why. Every volume must read on its own, wherever a reader starts.
Before. Cross-references used the three papers' ids (§1.3.1, T1.2.4.1, T2.5) and names (Main Paper, Technical Paper).
After. Every cross-reference uses the library's ids and names the document it points to when that is another one ('the Budget Book, B.7'; 'Follow the Plate, D.4.1'); references within a document print the id alone. The Executive Summary's connective lines name the Plan and the companion volumes; its page numbers are the Plan's. Old ids map to their new homes in src/layers/refs.py and src/layers/legacy-map.json, and every old web route redirects.
Why. No reference may send a reader to a document that no longer exists.
Before. A new major version is a new edition; the papers' structure or core changes is a major version.
After. A new major version is a new edition, made with contributors or reviewers beyond the initiator; a restructure by the initiator alone is a minor version. Document Control lists the whole library.
Why. Uncle James (28 Oct 2026): "keep it as v1.2; v2 should involve others beyond me." 1.2 is the first application of the rule.
1.1 27 October 2026 43 changes · shared version
First revision of 1.0. A review of who already helps hawkers beyond Government (hawkers' own associations, the labour movement, the five Mayors and CDCs, the People's Association and grassroots, Town Councils, charities, social enterprises, companies and researchers) and of overseas precedents, so that the plan builds on them and duplicates none. The Executive Summary gains a comparison of hawkers' protection with four other kinds of worker and a summary of the review; the retirement match now applies under 55, beside the existing scheme from 55. The pilot widens from one Lighthouse Centre to two Lighthouses from Year 1, Chinatown Complex and Geylang Serai, and Tekka from Year 2, five at most before the gate, and the programme is re-costed centre by centre from each centre's heritage-worthy stalls and size. Hawker-led Buying Circles (§2.12) and two asks of Government, a table-cleaning fee re-priced for tray return at the pilot centres and a named deputy for an hour a day, join the package, and the prep co-operative is dropped from the model as well. Earlier proposals are attributed even-handedly, each proposer named in the same parliamentary terms whatever the party, with what the plan agrees and disagrees with. A light first tier, the Heritage Watchlist (§3.6), opens at every government hawker centre from Year 1 and gives a yearly count of the stalls at risk; from Year 4 each centre funds new Fellowships at a maintenance rate fed by that count.
Before 1.2 the plan was three papers, the shared ancestors of both stalls; this entry is in both logs.
Approved by Uncle James · tag v1.1 · release commit b4fe4b7
Before. ES.1–ES.6.
After. New ES.3 The Thinnest Floor, and No Business to Sell (stallholder, assistant, platform worker, lower-wage employee, PMET) and ES.4 Who Already Helps, and Where This Plan Fits; the later sections renumbered ES.5–ES.8.
Why. Uncle James asked for the gaps hawkers face to be shown against gig, blue- and white-collar workers, and for the review of existing help to be summarised in the Executive Summary.
Before. §1.3 covered Government's rails only; §1.2 described FMAS as 'minority coverage, no leverage'.
After. New §1.3.1 Beyond Government: Who Else Already Helps (a table of 15 groups of actors and where the plan joins, defers or fills) and §1.3.2 What Only This Plan Does — and Where It Builds on Others (gaps no one fills; what is not new; overseas precedents). §1.2 describes FMAS fairly: wide reach, no welfare fund and no decision rights of its own.
Why. Make sure the plan has no blind spots and duplicates nothing already done by non-profits, charities, the Mayors and the People's Association.
Before. The retirement match applied at any age: base up to about $8.3M a year, stretch about $14.1M; shortfall over Years 4–10 $31.2M base, $55.5M stretch with HDB ($65.4M without); the endowment would need about $240M to carry the base.
After. The match applies to stallholders under 55; from 55 the Matched Retirement Savings Scheme already matches top-ups (up to $2,000 a year), and the plan helps hawkers claim it. Base at most about $4.1M a year, stretch about $8.0M (ceilings: fewer than half of stallholders are under 55); shortfall over Years 4–10 $6.8M base, $24.8M stretch with HDB ($29.8M without); about $120M of endowment would carry the base. Package programme cost about $80M (was $85.9M).
Why. The existing scheme already matches top-ups from 55; a second match would pay twice.
Before. The proposals did not name the existing actors working in the same space.
After. §2.5 asks that CDC targeted meal credits be spendable at hawker stalls, works through Project Belanja! and grassroots meal cards, and keeps the Affordability Compact funded, never contractual; §2.4 names the Foochow coffeeshop association; §2.7 recruits through Jobs Nearby @ CDC; §2.8 puts local pipelines first; §2.9 extends the FMAS–NTUC group cover; §2.10 adds cost data to price data others already publish.
Why. Build on what exists rather than beside it.
Before. These sections did not say what was already done or proposed.
After. §2.2 answers the 2024 parliamentary ballot proposal and the Government's reply; the Seed Bank's archive and apprenticeships acknowledge and partner existing documentation and training; the Hawker Council works with FMAS (a non-voting FMAS observer) and centre associations; the Fellowship and Taste Council build on the Hawkers Succession Scheme's stipend and tasting panel; the Heritage Plate is set apart from existing awards and labels; the Lighthouse is an NEA-managed centre; the Seed Bank registers under Arts and Heritage for the Cultural Matching Fund; the CDC Vouchers Donation Scheme is a giving route.
Why. Uniqueness and no duplication; each change follows the review in §1.3.1–§1.3.2.
Before. The food-stall assistant wage floor shown as an employee protection; the 2023 median wage ($5,070); the platform CPF rate without its age band; the Budget 2026 CPF Top-Up without its eligibility; about 6,900 cooked-food stalls attributed to NEA.
After. No wage floor binds a hawker stall; the 2025 median ($5,775, about 642 plates a day); platform and senior employee CPF rates by age; the Top-Up's eligibility and where it is paid; Jobseeker Support's $5,000 income cap; 6,900 marked as our estimate; a note splitting the employee column into the hawker's assistant, lower-wage employees and PMETs.
Why. Fact-check against 2025–26 official sources.
Before. —
After. 92 new sources for the review.
Why. Every new claim is sourced.
Before. ES.4 summarised the review in a list.
After. Fig. ES.1, a map of sixteen needs across a hawker's working life: who helps today (Government agencies, hawkers' own bodies, the Mayors and CDCs, the PA and grassroots, charities and others), what the plan does about each (fills a gap, builds on or joins), with its section, and who pays for the plan's part (the State, gifts and the endowment, or the trade itself). Drawn by src/layers/support_map.py.
Why. Uncle James asked for a visual map of existing support against the gaps the plan fills, with who pays as a further dimension.
Before. Fig. ES.1 showed hawkers' own bodies recognising masters at scale, and the plan building on it; ES.4 said they "recognise its masters".
After. Hawkers' own bodies recognise masters partly, and the plan reinforces that recognition (§4.2). ES.4: they "honour a few of its masters each year – with an award, not a wage", and the gaps include paying a master a year's wage to teach. §1.3.1 notes that the FMAS awards carry no published prize money, stipend or teaching role, that the Fellowship adds a year's pay to teach, and that FMAS's standing as the trade's voice is wide but not uncontested.
Why. Uncle James questioned whether hawkers' own bodies recognise masters sufficiently and meaningfully. They hold one ceremony a year with no published cash, so "at scale" overstated it and contradicted §4.2.
Before. Fig. ES.1 marked nine needs "Fills a gap", seven of them where some help already exists.
After. "Fills a gap" is kept for the two needs no one meets today (shared prep kitchens; repairs and exhausts). Retirement savings keeps "Fills, under 55". A new mark, "Reinforces it", shows the seven needs where help exists but is thin and the plan adds what it lacks: income once in a stall, days off, injury and illness cover, paying for goodwill, recipes kept to cook from, recognising masters and money that lasts. The legend gains a line for the plan's marks, and ES.4 says which rows are empty and which the plan reinforces. support_map.py now checks that "Fills a gap" appears only on rows where nothing was found.
Why. Uncle James pointed out that the plan reinforces areas the map shows as partly covered, and asked that the map say so.
Before. The 1.1 text drafted to a style sheet taken from the papers alone.
After. The same facts, re-voiced to Uncle James's own writing (his blog, essays and company updates).
Why. Uncle James asked that his writing style be studied from his own published writing.
Before. The plan proved everything at one Lighthouse Centre in Years 1–2. Year 3 was gate-dependent expansion: a second co-op site and the Security Package at a second centre, $600K in all. Expansion beyond the Lighthouse waited until the evaluator certified five gate criteria at month 20 or later (Fellowships ≥70%; first cohort landed, ≥80% operating at month 12; trust ≥55% and rising; three Succession Mortgages; the second endowment close). The rent-band pilot and the 50 Apprentice Work Passes were at the Lighthouse and its cluster.
After. One Lighthouse Centre opens in Year 1, as before. In Year 2 the plan widens to three pilot centres, five at most, alongside the Lighthouse. Each later centre passes its own §6.2 veto with its own hawkers willing, gets its trust deposits (the Security Package, listening and co-design) in the six months before it opens, in Year 1, and runs the full loop: Master Fellowships, apprentices at the masters' own stalls, a Relief-Corps bench, the Security Package, the rent-band and ballot pilot and evaluation. The Taste Council, the repository and the core institution stay national and shared. No further than five until the evaluator certifies the five gate criteria, now measured across all the pilot centres at month 43 or later after the Lighthouse opens (Fellowships ≥70% with no centre below 50%; every centre's first cohort landed, ≥80% operating at month 12; trust ≥55% and rising at every centre; at least six Succession Mortgages at two centres or more; the final close, with each further centre's first two years funded). The gate governs going beyond five, from Year 4, not a Year 3 second site. The rent-band pilot and the 50 passes run across the pilot centres and their clusters.
Why. Uncle James (27 Oct 2026): "We need more data points than just one Lighthouse Centre. Not every hawker is going to be willing to try our schemes out."
Before. Four-year programme $9,125,460 (Year 0 $711,700; Year 1 $2,358,020; Year 2 $2,467,420; Year 3 $3,588,320, of which $600K gate-dependent): two co-ops, 18 Master Fellowships, 36 apprenticeships, 13 landings. From Year 4, $2,988,320 a year, indexed. Gaps of about $3.1M across Years 1–3. No later year short; the corpus kept its real value in about 19 futures in 20. Retirement-match shortfall over Years 4–10 $6.8M base, $24.8M stretch with HDB ($29.8M without). Downside: Lighthouse-only Year 3 roughly $2.9M. About $1.61 of public money for every private dollar in the campaign. First-loss budget $410,000.
After. At three centres, the central plan, $12,521,650 (Year 0 $526,900; Year 1 $2,215,990; Year 2 $4,272,430; Year 3 $5,506,330), after the 1.1 removals below: 38 Master Fellowships, 76 apprenticeships, 23 landings. Each further centre, the fourth and the fifth, is a stated add-on of $2,570,140 ($196,900, $994,770, $1,378,470); five centres $17,661,930. From Year 4, $5,506,330 a year, indexed, and $1,378,470 more for each further centre. Gaps of about $6.7M across Years 1–3. The programme spends every continuing gift and is $1.7M short over Years 4–10 ($10.9M over Years 4–20); the corpus keeps its real value in about half of futures (52 in 100). Retirement-match shortfall $15.7M base, $33.6M stretch with HDB ($38.6M without). Downside: the Lighthouse alone, $2,663,590 in Year 3. About $1.37 of public money for every private dollar. First-loss budget $570,000.
Why. Uncle James: "Let the budget rise, and say so." Each later pilot centre's lines are the Lighthouse's own, one year behind, with its deposits in Year 1; a shared pilot-centre coordinator, per-centre evaluator, Taste Council and events lines, and the 10% contingency. Revenue is scaled per centre, at what the Lighthouse earned at the same age. Every new assumption is labelled ASSUMPTION in the funding model's notes (T1.4).
Before. The 5,000 futures varied whether the gates opened for Year 3's $600K expansion (70%).
After. They vary whether each of the two later pilot centres passes its hawkers' vote (80% each; a centre voted down keeps its Year 1 deposits and costs nothing after). Four and five centres, and the Lighthouse alone, are run as scenarios.
Why. The gate no longer governs any Year 3 spend; a centre's own veto is the uncertainty the plan cannot control.
Before. The Hawker Security Package gave three deposits at each pilot centre before any ask: group critical-illness cover for stallholders 55+, an exhaust and ducting deep-clean blitz across ~120 stalls ($78,000), and a same-week equipment repair fund ($90,000), with yearly renewals of insurance and repairs ($105,000–$145,000 a centre). The Package was billed as the successor to the Hawkers' Productivity Grant. Fig. ES.1 showed "Repairs and exhausts" as a gap the plan fills.
After. The Package is cover only: the group critical-illness pool for stallholders 55+ ($52,500 a centre a year, about 70 at about $750), and the FMAS–NTUC group-cover route widened to critical illness and to stallholders. Exhausts, ducting and stall repairs are the landlord's (NEA or the Town Council) and the hawker's own business, and the papers say so. Fig. ES.1 keeps the need, marked "Left to landlords", with no section and no payer.
Why. Uncle James (27 Oct 2026): "Why are you budgeting on infrastructural items of the hawker centre? That's not our programme objectives." The savings are banked, not re-spent.
Before. §2.6 Prep Co-operatives — Twin Certified Lines: a hawker-owned co-op kitchen at each pilot centre with halal and non-halal lines (fit-out $420,000 at the Lighthouse, $380,000 at each later centre; operations $60,000–$85,000 a year; member fees $150–$300 a month; a co-op manager in the core team), claiming a 14–16-hour day compressed towards 10–12. Programme-run heritage stalls staffed by apprentices (2, 3, then 5 at the Lighthouse; 2 then 3 at each later centre; setups, rent subsidy and stall net). The incumbent package in Follow the Plate included the co-op.
After. §2.6 Learning at the Master's Side – Nothing to Build: apprentices learn at the master's own stall as registered stall assistants (building on the Hawkers Succession Scheme), each record is proved on the master's relief day, and blind tastings and retired masters' documentation book partner kitchens by the session; graduates land through the rent-band ballot or the Incubation Stall Programme. The only costs are a one-off $3,000 second-station grant at each new Fellow's stall and about 24 kitchen sessions a centre a year at $300 (both ASSUMPTIONS). No co-op, no programme-run heritage stalls, no co-op manager; the plan no longer claims a shorter prep day. Follow the Plate keeps the co-op only as a tested variation, outside the package: with the package a tendered stall stays above the assistant benchmark up to about $1,660 (was $2,280), and 41–55% of tendered stalls fall below it across $1,200–1,800 (was 28–41%).
Why. Uncle James (27–28 Oct 2026): a working hawker will not leave his stall to prep elsewhere and come back; shared prep only makes sense for cooking recorded dishes to spec, which is training. He chose an asset-light design that takes no stalls (approved 00:15 SGT, 28 Oct).
Before. Each year's ledger was one list of lines, at gross cost, with a 10% contingency, programme revenue and a net requirement.
After. Each year's ledger has two parts: A, the operating budget (the institution's burn: core team, pilot-centre staff and liaison officers, interim lead and research assistants, office, legal, audit and insurance, communications, the evaluator, Hawker Council fees and the shared national platforms), and B, the programme budget (grants and expenses that reach hawkers), shown per centre for the Lighthouse and pilot centres 2 and 3, each with its contingency, revenue deducted from B; the net requirement is A + B. Each year prints the share reaching hawkers directly (66.0% over Years 0–3) and the operating burn a month ($88,800 on average). §7.1 adds a summary of the split and a budget for one pilot centre. The funding model carries an operating and a per-centre programme line (funding.js in parity).
Why. Uncle James asked to split the programme staff budget from the hawker programme grants and expenses, "like a tech startup".
Before. Follow the Plate kept the prep co-operative as a tested variation outside the package ($225 a month, ingredients 2% and utilities 10% cheaper, the day from 13 to 11 hours): T1.2.7 scored it (leverage 0.22, 3.57 with time), T1.2.5.2 and T2.4 described it, and the simulator offered it.
After. The co-op is gone from the model (plate_model.py and its browser port, in parity), the simulator, the site's Follow the Plate page and the Technical Paper. The Main Paper keeps only §2.6's account of why it was dropped.
Why. Uncle James (28 Oct 2026): drop the prep co-op completely, including the tested variation.
Before. No proposal helped hawkers buy together; the papers called buying together a gap.
After. New §2.12, The Buying Club – Buy Together, Own Nothing: members buy ingredients on negotiated wholesale contracts, delivered to the stall, through software with no warehouse; one power contract on the open electricity market, with gas through the same club; delivery at a 15% group commission or through a commission-free channel building on Hawkers United – Dabao 2020. ASSUMPTIONS: ingredients 3% and utilities 10% cheaper, commission 15% (was 22%). Self-funding with no member fee: a 1% supplier rebate on about $40,000 a member a year (ASSUMPTION) against a procurement and partnerships lead (1 FTE, $78,000) and ordering software ($30,000, then $18,000 a year) in the operating budget; it pays for itself from about 265 members, and at three centres the operating budget carries $216,000 over four years. It concedes operators' bulk buying, little used (MSE, 10 Jan 2024), and builds on it without a new agency. In the package.
Why. Uncle James (28 Oct 2026): add asset-light levers, all of them, to the package; costs in the ledgers, the role in the operating budget.
Before. §2.2 treated rent alone as the occupancy cost; the table-cleaning fee (2025 median $520 a month) was not discussed.
After. §2.2 adds the fee beside the rent: an ask of NEA and operators to re-price the table-cleaning fee for the work left since tray return became the rule in 2021. ASSUMPTION: $200 a month less, at NEA and operator-run centres. The model books it as income the fee collector gives up (landlord share), an upper bound: about $16.6M a year across 6,900 stalls, $16.0M of it the State's and $0.5M the operators', counted as programme cost; shown in T1.2.2.2, T1.2.5, T1.2.7.1 and T1.2.7.4, and as a row in §7.7.
Why. Uncle James (28 Oct 2026): the fee was set before tray return; ask Government and operators to re-price it, and show who bears it.
Before. The relief notice was the only change asked of the personal-operation rule.
After. §2.7 adds a named deputy, an NEA rule change: one named, registered assistant may run the stall for up to an hour a day, logged like the relief notice, while the stallholder stays on the roll. It concedes that the personal-operation rule exists to stop subletting and sets out the safeguards. ASSUMPTION: the hawker works 12 of his 13 hours, at no extra wage. Zero fiscal cost; a row in §7.7. In the package.
Why. Uncle James (28 Oct 2026): a named deputy, an ask of Government, with safeguards so it cannot become subletting.
Before. Package (price freedom, Makan Vouchers, Relief Corps, Protection Floor): a typical tendered stall above the $11.64 assistant benchmark up to about $1,660 of rent; 41–55% of tendered stalls below it across $1,200–1,800 with the package (62–75% today); $12.28 an hour with the package; national package programme cost about $77.4M a year, leverage 0.90 (cash 0.79).
After. Package with the buying club, the fee reset and the named deputy: above the benchmark up to about $2,590 (about $2,300 without the deputy); 22–34% below it across $1,200–1,800 (62–75% today, unchanged); $15.43 an hour; national package programme cost about $93.4M a year with the fee reset counted in full, leverage 1.12 (cash 1.01). Each lever alone on the first four: ingredients $1,900, utilities $1,760, fee $1,860, delivery $1,760, deputy $1,950. The successor's protection still binds from about $1,920 and a stall still loses money only above about $4,500.
Why. Recomputed with the new package.
Before. $12,521,650 over Years 0–3 at three centres (Y1 $2,215,990, Y2 $4,272,430, Y3 $5,506,330); operating $4,263,160, 66.0% reaching hawkers, burn $88,800 a month; each further centre $2,570,140; Lighthouse alone $7,123,970 ($2,663,590 in Year 3); bridge grants $6.7M; from Year 4 about $0.24M a year short ($1.7M over Years 4–10).
After. $12,737,650 (Y0 $526,900, Y1 $2,322,790, Y2 $4,336,030, Y3 $5,551,930); operating $4,479,160, programme unchanged at $8,258,490, 64.8% reaching hawkers, burn $93,300 a month; each further centre $2,540,140; Lighthouse alone $7,399,970 ($2,745,190 in Year 3); bridge grants $6.9M ($1.6M for the Lighthouse alone); from Year 4 about $0.29M a year short ($2.0M over Years 4–10); about $1.36 of public money for every private dollar in the campaign.
Why. The buying club's lead, software and supplier rebates in part A of each ledger.
Before. Fig. ES.1 mapped fifteen needs.
After. Fig. ES.1 maps sixteen: 'Buying ingredients and power' under Running the stall, marked as partly helped by others (some centre operators offer bulk buying, little used: MSE, 10 Jan 2024), which the plan reinforces (§2.12), paid for by the trade. Repairs and exhausts remain the only empty row, left to landlords.
Why. Uncle James (28 Oct 2026) asked for the row; the sources show some help exists today, so it is not marked 'nothing found'.
Before. 1.0 did not mention the earlier proposals. The first 1.1 drafts named the party behind the motion of 13 November 2024 but the Member who asked the question of 14 October 2024 without his, and said the plan 'takes the idea' of centralised procurement; a later draft named no party or Member and deleted the procurement mention. §1.3.2 still spoke of a one-centre pilot.
After. Even-handed attribution. Every proposer is named in the same parliamentary terms, with no adjectives: "Leong Mun Wai, a Non-Constituency MP of the Progress Singapore Party" moved the motion; "Melvin Yong, the People's Action Party MP for Radin Mas" asked the question; "Hazel Poa, a Non-Constituency MP of the Progress Singapore Party" proposed centralised procurement through a new agency; the House passed the motion as amended by "Edward Chia, the People's Action Party MP for Holland–Bukit Timah". The Government's answers stay. The plan says what it agrees with and what it does not, and judges the idea, not the proposer: "I agree that stalls should be allocated by ballot rather than by the highest bid. I do not agree with a flat rent" (§1.3.2, §2.2); "I share the aim of cheaper ingredients for hawkers. I do not agree with a single national buyer, even an optional one: it would narrow the competition among Singapore's suppliers, not widen it" (§1.3.2, §2.12). "$500 or 3% of Gross Turnover, whichever is higher" and "a sixth category, namely heritage foods" are quoted from the mover's own speech and marked as his words. References: the two speeches restored, marked as published by the party and cited only for the speakers' own words, beside MSE's response to the motion and its oral reply of 14 Oct 2024; the sources data gains both speeches and the oral reply. §1.3.2 now says three pilot centres, and the §7.7 rent-bands row says the plan "keeps the ballot proposed in Parliament in 2024, not its flat rent".
Why. Uncle James (28 Oct 2026, 07:34): "You can mention centralized procurement and you can note the ballot motion and earlier parliamentary questions. Don't avoid it. I just don't want us to be seen to be endorsing ideas they originate as if we agree with them. Make it clear what we agree with or not, but don't show partiality."
Before. The model note for the Lighthouse-alone operating budget said: "Years 0–2 are the Lighthouse's ledgers as first costed, less the exhaust blitz, the repair fund and the repairs in each renewal."
After. The stale sentence is deleted; the ledgers no longer carry hawker-centre infrastructure, so there is nothing to subtract. The figures are unchanged.
Why. Uncle James (28 Oct 2026) asked for the stale note to go.
Before. §2.12, The Buying Club – Buy Together, Own Nothing: a members' club run in software by the Seed Bank, negotiating wholesale contracts and one power contract, self-funding through a 1% supplier rebate against a procurement lead (1 FTE, $78,000) and ordering software; $216,000 over four years at three centres.
After. §2.12, Buying Circles – Hawkers Bargain Together. The Seed Bank never buys, stocks, sets prices or picks suppliers, and never takes title to goods or money; every hawker keeps his own supplier and recipe. Through their own hawkers' association the stalls at a centre form small circles (rice, oil, poultry, gas), voluntary, with no exclusivity, free to leave any week; a circle lead, usually a hawker, receives a small honorarium (ASSUMPTION: $100 a month). Circles post their pooled weekly volume on an open price board and suppliers compete; operators' bulk-purchase offers join the board. Power by group switch on the open electricity market, gas by circle, delivery rates negotiated per centre or commission-free channels. The Seed Bank provides the board (on the Plate Ledger's data), group-order templates and one part-time partnerships role (0.5 FTE, $39,000 a year); the full procurement FTE, the software and the supplier rebates go. Savings unchanged and marked ASSUMPTION (ingredients 3%, utilities 10%, commission 22% to 15%). No member fee; the circles do not pay for themselves at three centres, and the operating budget carries $185,460 over four years.
Why. Uncle James (28 Oct 2026): "We can't be over-centralising procurement into a single national agency; that's akin to nationalising hawker procurement, which can be sensitive and sends wrong signals to supplier entrepreneurs."
Before. The table-cleaning fee reset was asked for at every NEA and operator-run centre and booked nationally, up to about $16.0M a year to the State and $0.5M to operators, inside the national package cost of about $93.4M a year; ES.4 counted four places where the State is asked for money.
After. The reset is asked of NEA at the pilot centres only, three, five at most, so the evaluator shows its effect before anyone prices it nationally. ASSUMPTION: a pilot centre has about 120 stalls, like the Lighthouse, each paying the fee: up to about $864,000 a year at three centres ($288,000 a centre), $1.44M at five, an upper bound. The national figure (about $16.6M a year, $16.0M the State's) is shown apart as a later decision, outside the package total and the plan. A pilot-centre stall still gets the reset, so the package's per-stall effect and the break points are unchanged. ES.4 now counts five places: four Budget lines and the fee income given up at the pilot centres.
Why. Uncle James (28 Oct 2026): ask for the reset at the pilot centres first; present the national figure only as an upper bound for a later decision.
Before. $12,737,650 over Years 0–3 at three centres (Y1 $2,322,790, Y2 $4,336,030, Y3 $5,551,930); operating $4,479,160, 64.8% reaching hawkers, burn $93,300 a month; each further centre $2,540,140; Lighthouse alone $7,399,970 ($2,745,190 in Year 3, gaps $1.6M); from Year 4 about $0.29M a year short ($2.0M over Years 4–10); national package programme cost about $93.4M a year, leverage 1.12 (cash 1.01).
After. $12,707,110 (Y0 $526,900, Y1 $2,275,170, Y2 $4,335,570, Y3 $5,569,470); operating $4,448,620, programme unchanged at $8,258,490, 65.0% reaching hawkers, burn $92,700 a month; each further centre $2,580,700 (82% reaching hawkers); Lighthouse alone $7,288,310 ($2,716,170 in Year 3, gaps $1.5M); bridge grants still about $6.9M; from Year 4 about $0.31M a year short ($2.2M over Years 4–10); about $1.36 of public money for every private dollar in the campaign, unchanged. National package programme cost about $78.3M a year with the fee reset at the pilot centres only, leverage 1.14 (cash 1.03). Break points unchanged.
Why. Recomputed with the Buying Circles' costs in part A and the fee reset at the pilot centres only.
Before. T3.2's funding-model notes still described how the ledger was derived from the old prep co-op lines (the earlier combined Taste Council and equipment line, the co-op manager, the co-op site, the split revenue lines).
After. Those derivation notes are deleted; the notes describe the ledger as it now stands.
Why. Uncle James (28 Oct 2026): delete stale notes.
Before. The Lighthouse and the pilot centres were to be centres NEA manages itself, so that one landlord answered for every mechanism; Town Councils deferred (§1.3.1), and the fee reset was an ask of NEA alone, booked as income the State gives up at NEA centres.
After. Any of the 123 government hawker centres can host a pilot centre: 27 NEA-managed (26 open), 80 Town Council-managed and 16 run by socially-conscious enterprise operators (NEA list, 17 Aug 2026). NEA regulates every tenancy and stays the tenancy landlord at the 107 NEA and Town Council centres, so rents, bands, the ballot and tenancies work the same there. A Town Council joins as a partner: it manages the common property and collects the conservancy and table-cleaning fees, so the fee reset at its centre needs its agreement and the income given up is its own. An operator also sets tenancy terms within NEA's framework, so the bands, the ballot, its fees and the reset need its agreement, and any share of fee income it gives up is its own, not the State's. Mixing managers tests the plan under each. Privately owned coffeeshops and food courts are left out for now; one may join only on the board's recommendation and without the Hawker Council's veto (§3.8, §6.2). The fifth place the State is asked for money is now the fee income NEA or a Town Council gives up at the pilot centres. No budget figure or stall-count assumption changes.
Why. Uncle James (28 Oct 2026, 09:42 and 09:53 SGT): "I won't stick to rules that don't comprehensively represent hawker centres; expand the definition to include Town Council-managed ones"; then include the 16 operator-run centres, which are government centres, and exclude only privately owned coffeeshops and food courts.
Before. §6.3 described the Lighthouse as a mid-size, mixed-cuisine centre with a substantial 55+ cohort and named no centre.
After. Seven selection tests: a government centre; footfall; heritage density (MICHELIN Guide or Hawker Heritage Award stalls, stalls of 30+ years or multi-generation); between them Chinese, Malay and Indian cuisine, with halal stalls; a mix of NEA and Town Council management; no major works in the pilot window; the association's support and the §6.2 vote. Geylang Serai Market and Food Centre is included by default, subject to its hawkers' vote; if they decline, the plan seeks another centre strong in Malay cuisine. Examples to consider, each subject to its own vote: Chinatown Complex (a possible first centre), Tekka Centre, Old Airport Road (Blk 51) and Maxwell Food Centre. The paper notes that Chinatown Complex and Old Airport Road are bigger than the 120-stall centre the ledgers cost. Twelve new sources.
Why. Uncle James (28 Oct 2026): "Describe the criteria while naming possible locations as examples to be considered"; "consider the importance of Malay cuisine, so we are racially balanced and inclusive; by default at least include Geylang Serai Market and Food Centre."
Before. One Lighthouse Centre opened in Year 1 and two more pilot centres a year later, Geylang Serai included by default and Chinatown Complex, Tekka, Old Airport Road and Maxwell named as examples. §6.3 was titled "The Lighthouse and the Pilot Centres" and §7.3 "Year 1 — The Lighthouse Launches".
After. Two Lighthouses open side by side in Year 1, Chinatown Complex and Geylang Serai Market and Food Centre, for racial balance and inclusion (Chinese and Malay cooking, halal stalls from the start); Tekka Centre is the third pilot centre and opens in Year 2. Old Airport Road (Blk 51) and Maxwell are examples of a fourth and a fifth, stated add-ons. Every centre is subject to its own hawkers' vote; if one declines, the plan seeks another centre with a similar cuisine strength. Year 0 trust deposits, listening and co-design run at both Lighthouses, and Tekka's in the six months before it opens, in Year 1. The gate is unchanged: no further than five centres until the evaluator certifies the criteria across all the pilot centres, its month 43 counted from when the Lighthouses open. One term throughout: "the two Lighthouses", "the Lighthouses", "a Lighthouse"; Part VI is "Year Zero & the Lighthouse Centres", §6.3 "The Lighthouses and the Pilot Centres" and §7.3 "Year 1 — The Lighthouses Launch", with site redirects from the old routes.
Why. Uncle James (28 Oct 2026): two Lighthouse centres in Year 1, Chinatown Complex and Geylang Serai, side by side for racial balance and inclusion; Tekka the third centre, in Year 2; Old Airport Road and Maxwell as add-ons; rename consistently.
Before. Each later pilot centre was assumed to be a mid-size centre of about 120 stalls and about 70 stallholders aged 55+, at the Lighthouse's ledger rates one year behind: 18 Master Fellowships at the Lighthouse (4, 6 and 8) and a flat 10 at each later centre (4, then 6), 20 apprentices, about 13 dishes and 5 landings; each further centre added $2,580,700.
After. Each centre is costed from its own shape. Heritage-worthy stalls, a proxy counted on 28 Oct 2026 (Tier 1: MICHELIN Bib, Selected or Plate stalls, current or former stars, Hawker Heritage Award winners; Tier 2: stalls documented as running 30+ years, or second generation or later; each counted once), padded for guide coverage into a planning count: Chinatown Complex 226 cooked-food stalls, 18 counted, x1.5, 27; Geylang Serai 63, 7, x1.2, 8; Tekka 119, 6, x1.5, 9; Old Airport Road 168, 11, x1.5, 16; Maxwell 103, 13, x1.0, 13. One Master Fellowship ($72,000) per stall in the planning count over the centre's three years, a ceiling, unspent money returning to the endowment, phased 30/35/35% (Year 1 centres) or 40/60% (Year 2 centres): 27 (8, 9, 10), 8 (2, 3, 3) and 9 (4, 5). Two apprentices a Fellowship ($18,000 a year), one documented dish a Fellowship ($18,000), landings for about 30% of apprentices, a second-station grant ($3,000) and $15,000 of first-loss reserve and escrow a Fellowship (the former $570,000 for 38). By stall count: the critical-illness pool (70 in 120 stallholders, $750: 132, 37 and 69), the Corps bench (about 10 cooks a 100 stalls, two-thirds in the first year, $40,000 a bench plus $12,500 a member: 15 then 23, 4 then 6, 8 then 12), circle leads (four a centre at least, one for about every 30 stalls: 8, 4 and 4) and centre staff (1.5 FTE for every 120 stalls, rounded down to the half, at least 1.5; Chinatown Complex adds one beyond the core team's Lighthouse roles). Revenue $10,000, $14,500 and $25,000 a Fellowship by the centre's age. Each Lighthouse buys a $30,000 capture kit and pilot-documents three hawkers in Year 0. A new T1.4.6 sets out the proxy, its padding and its sources (three new).
Why. Uncle James (28 Oct 2026): "consider the shape of each hawker centre; how many are likely to be heritage/conservation worthy? Use proxy data, e.g. popularity ... as proxy inputs ... with some padding as budget-per-centre logic."
Before. $12,707,110 over Years 0–3 (Y0 $526,900, Y1 $2,275,170, Y2 $4,335,570, Y3 $5,569,470); operating $4,448,620, programme $8,258,490 (65.0% reaching hawkers); burn about $92,700 a month; 38 Fellowships, 76 apprenticeships, 23 landings; bridge grants for gaps of about $6.9M; from Year 4 about $5.65M a year, $0.31M a year short ($2.2M over Years 4–10, $12.1M over Years 4–20); corpus keeps its real value in 50% of futures; $1.36 of public money per private dollar; downside, the Lighthouse alone, $2,716,170 in Year 3 ($7,288,310 over four years); first-loss budget $570,000.
After. $14,628,275 (Y0 $824,725, Y1 $3,479,950, Y2 $4,895,500, Y3 $5,428,100); operating $4,910,840, programme $9,717,435 (66.4%); burn about $102,300 a month; 44 Fellowships, 88 apprenticeships, 26 landings, 238 stallholders 55+ covered; gaps of about $8.5M ($2,459,950, $3,045,500, $3,028,100), Year 0 foundation grants $631,725; from Year 4 about $5.43M a year in Year 3 prices ($5.51M in Year 4), $0.16M a year short ($1.1M over Years 4–10, $9.4M over Years 4–20); corpus keeps its real value in 46% of futures; $1.27 of public money per private dollar. Add-ons: Old Airport Road $3,888,980, Maxwell $3,072,400 (four centres $18,517,255, five $21,589,655); a further centre not yet named (100 stalls, 10 heritage-worthy) $2,587,250. Downside: both Lighthouses keep running and Tekka defers, $4,135,775 in Year 3 ($12,138,450 over four years), no later year short. First-loss budget $660,000 ($560,000 after escrow set-up). The Monte Carlo now draws Tekka's veto alone (80%); a Lighthouse that declines in Year 0 is replaced.
Why. Recomputed from the per-centre ledgers (funding_model.py, in parity with funding.js).
Before. The fee reset at the pilot centres assumed about 120 stalls a centre, about $0.86M a year at three ($288,000 a centre) and $1.44M at five; the national package about $78.3M a year (cash leverage 1.03).
After. Counted at each centre's actual cooked-food stalls, $2,400 a stall a year: Chinatown Complex $542,400, Geylang Serai $151,200, Tekka $285,600, $979,200 at the three (408 stalls) and $1,629,600 at five (679 stalls, with Old Airport Road and Maxwell); the national package about $78.4M a year (leverage 1.14, cash 1.02).
Why. Uncle James (28 Oct 2026): lines that follow the stall count use each centre's actual count.
Before. No comment on how the flat allocation compared with each centre's heritage.
After. "The flat allocation was about right in total – 38 Master Fellowships, 18 at the one Lighthouse and 10 at each later centre, against the 44 the heritage count asks for – but in the wrong places: nine short at Chinatown Complex, two too many at Geylang Serai and one at Tekka. Yet money is not the real limit. Willing masters are: NEA's Hawkers Succession Scheme completed six handovers between January 2022 and September 2025 (MSE, 22 Sep 2025). So the Fellowship line is a ceiling, not a target, and what a centre does not spend on it returns to the endowment."
Why. Uncle James (28 Oct 2026): say that the flat allocation was about right in total but in the wrong places, and that willing masters, not money, are the real limit.
Before. §3.6 Endangered Dish Registry, a ranked register of dishes kept with NHB, culinary historians and communities; no stall-level tier and no yearly count.
After. A light first tier, the Heritage Watchlist, national from Year 1 at all 123 government hawker centres. A stall qualifies on any one test (20 years or more at the stall; a second generation or later; MICHELIN, the Hawker Heritage Award or an NHB or NEA heritage board; or a nomination by its hawkers' association or by 20 regulars, an assumption), confirmed by one short visit; private coffeeshops and food courts by the existing exception route. A listed stall gets a public listing and a counter decal ("On the Heritage Watchlist"; recognition, not a quality mark, so it does not compete with the Hawker Heritage Award or the Heritage Plate), a story-and-dish capture for the open repository and first call on Relief-Corps days. A yearly check-in with consent asks four questions (age band, named successor, tenancy end, health); only aggregates are published, in line with the PDPA. Two or more signals (owner 65 or over, no successor, tenancy ending within three years, among the last stalls cooking the dish) flag a stall at risk: it goes to the front of the queue for a Master Fellowship and full documentation, and its dish feeds the Registry. Each year the count of stalls watchlisted and at risk, by centre, cuisine and community, goes to NEA and MSE, NHB and MCCY, and the Singapore Tourism Board, from the end of Year 1; §7.7 adds a zero-cost row asking them to use it. Assumed 300, 600 and 800 stalls by the end of Years 1–3 (12–15% of the 6,000+ cooked-food stalls), 15–25% of them at risk (about 120–200 by Year 3). Cost in the operating budget, because its stipends pay visitors, not hawkers: $99,000, $107,000 and $114,000 in Years 1–3 before contingency (a 1 FTE coordinator; 1,700 visits at a $30 stipend and $10 travel; $10 decals; the listing on the open price board, the increment only), $352,000 with contingency.
Why. Uncle James (28 Oct 2026): "a lite version to pre-qualify a stall as heritage worthy so we can encapsulate more stalls … you're in my watchlist", so the plan can move quicker to the stalls truly at risk and give Parliament and its partners (NEA, the Singapore Tourism Board and others) an at-risk count.
Before. §3.6 Endangered Dish Registry.
After. §3.6 The Heritage Watchlist and the Endangered Dish Registry, not renumbered; the Watchlist first, as the stall-level front door, and the Registry second, as the dish-level ranking. Its old web route redirects.
Why. Uncle James (28 Oct 2026): the Watchlist is the first tier of §3.6.
Before. Fig. ES.1 mapped sixteen needs; one row was empty (repairs and exhausts, left to landlords). ES.2's sixth cell was left empty.
After. Fig. ES.1 maps seventeen: "Knowing who is at risk" under Handing over, keeping the dish, with nothing found today (MCCY, 7 Oct 2026, keeps no list; NHB's inventory of intangible cultural heritage names practices, not the stalls at risk), and the plan filling the gap (§3.6, paid from gifts and the endowment). ES.4: two rows are empty; the plan fills one, knowing who is at risk, and leaves the other to the landlords. ES.2: the Watchlist and the Registry would fill the sixth cell.
Why. Uncle James (28 Oct 2026): show the new need in the map, honestly marked.
Before. $14,628,275 over Years 0–3 (Y0 $824,725, Y1 $3,479,950, Y2 $4,895,500, Y3 $5,428,100); operating $4,910,840, programme $9,717,435 (66.4%); burn about $102,300 a month; gaps of about $8.5M; listening fees 150 at every centre; Old Airport Road $3,888,980, Maxwell $3,072,400; the two Lighthouses alone $12,138,450 ($4,135,775 in Year 3); $1.27 of public money per private dollar.
After. $14,990,340 (Y0 $834,955, Y1 $3,588,685, Y2 $5,013,200, Y3 $5,553,500); operating $5,262,840, programme $9,727,500 (64.9%); burn about $109,600 a month; gaps of about $8.9M ($2,568,685, $3,163,200, $3,153,500), Year 0 foundation grants $641,955. Listening participant fees follow the stall count, about 1.25 a cooked-food stall (150 for the 120-stall centre the ledgers were first drawn for): 283 at Chinatown Complex, 79 at Geylang Serai (was 150 for 63 stalls), 149 at Tekka. Old Airport Road $3,898,880, Maxwell $3,068,935 (four centres $18,889,220, five $21,958,155); a further centre $2,583,125. The two Lighthouses alone $12,500,680 ($4,261,175 in Year 3). $1.26 of public money per private dollar.
Why. Uncle James (28 Oct 2026): the Watchlist's cost, and the open item that Geylang Serai's listening carried 150 participant fees for 63 stalls.
Before. From Year 4 each centre kept its Year 3 budget, indexed, so Chinatown Complex funded ten new Fellowships a year beyond its 27 heritage-worthy stalls; about $5.51M a year at three centres, $0.16M a year short ($1.1M over Years 4–10); every continuing gift spent; the corpus kept its real value in 46% of futures and ended Year 20 at 86% of target; retirement-match shortfall over Years 4–10 $15.7M base, $33.6M stretch with HDB ($38.6M without).
After. From Year 4 the steady state follows the heritage logic: once a centre's planning count has been offered, new Fellowships continue at a maintenance rate fed by the Watchlist's at-risk flow, an assumption of about 10% of the planning count a year, at least one (3 at Chinatown Complex, 1 at Geylang Serai, 1 at Tekka; 2 at Old Airport Road, 1 at Maxwell), with their apprentices, dishes, grants, reserve and landings; stall-driven lines and the national core stay at their Year 3 level. About $3.50M a year at three centres in Year 3 prices ($1.02M, $0.51M, $0.62M; add-ons $0.89M and $0.58M), against about $5.31M from the draw, the operating match and continuing gifts: no year short, a little under half of continuing gifts spent, the corpus keeps its real value in 89% of futures and ends Year 20 at 160% of target. At 20% a year it ends at 132%, at 30% at 103%, and at 40% three centres run $0.8M short over Years 4–10. The two Lighthouses alone $2.88M a year from Year 4. Retirement-match shortfall over Years 4–10 with half of surplus gifts set aside: $8.8M base (short from Year 7), $26.8M stretch with HDB ($31.7M without); the corpus then ends Year 20 at 118% of target.
Why. Open item from the last pass: the Year 3 envelope kept funding Fellowships beyond each centre's heritage-worthy stalls; Uncle James (28 Oct 2026) asked for the steady state to follow the heritage logic, fed by the Watchlist.
Before. The Executive Summary's cover said "in seven pages", but the 1.1 additions run to eight.
After. Uncle James accepted eight pages (28 Oct 2026). The cover now says "in eight pages"; ES.7's four figure tiles sit on one row.
Why. The cover states the length the summary actually has.
1.0 27 October 2026 1 change · shared version
First edition. Hawker Heritage is published as three papers that share one version: the Executive Summary (why and what), the Main Paper (what and how) and the Technical Paper (how it holds and how to vary it). This entry is the baseline against which every later change is recorded.
Before 1.2 the plan was three papers, the shared ancestors of both stalls; this entry is in both logs.
Approved by Uncle James · tag v1.0 · release commit 4ada8d6
Before. —
After. The full text of all three papers as released on 27 October 2026.
Why. First release to readers.
Stall 02
Saving our Hawker Heritage
Version 1.11 · 2 November 2026 · tag plan-v1.11
1 · Executive Summary · 2 · The Plan · B · The Budget Book. 15 versions and 357 changes recorded.
↓ Download 0P · the Plan Change Log (PDF, 522 KB, 215 pages) The Plan
The PDF also holds the development history before 1.0 and the SHA-256 of every released PDF.
1.11 2 November 2026 2 changes
Its own version. Saving our Hawker Heritage now has its own version, history, tag and Version and Change Control Log (0P, the Plan Change Log), starting at 1.11. Until 1.10 it shared one version and one log with The Hawker Library. On the website, every Plan page now lives under /heritage/. No proposal, figure or claim changes.
Approved by Uncle James (2 Nov 2026) · tag plan-v1.11
Before. One version for all twelve documents, 1.10, tagged v1.10, and one Version and Change Control Log, 0 (Hawker_Heritage_0_Change_Log.pdf), for both stalls
After. Saving our Hawker Heritage has its own version, 1.11, tagged plan-v1.11, and its own log, 0P, the Plan Change Log (Hawker_Heritage_0P_Plan_Change_Log.pdf). Each of its documents prints that version on its cover, in its footer, in its metadata and on its Document Control page, which now names the stall and its log. The shared history, 1.0 to 1.10, is in both logs: each row of it sits in the log of the stall whose documents it touched, and in both when it touched both. The website's /changes/ page shows both logs
Why. Uncle James's decision of 2 Nov 2026: each stall keeps its own version and change log, so a change to one stall no longer moves the other's version. A restructure by the initiator alone, so a minor version
Before. "the twelve documents", "all twelve documents", "one of the twelve documents" (since 1.10, for every document of Hawker Heritage with its one change log)
After. "the thirteen documents", "all thirteen documents", "one of the thirteen documents": the eleven documents and the two change logs, 0L and 0P
Why. With a change log for each stall there are thirteen documents, not twelve; no other word changes
1.10 2 November 2026 6 changes · shared version
Two stalls. The documents now sit in two series, the Hawker Library (the evidence) and Saving our Hawker Heritage (the Plan), each by Uncle James, a Silicon Wok programme. The Library adds a graded estimate of what hawkers earn above the median, and the Plan uses it in three places. No proposal, ask or model figure changes.
Approved by Uncle James (2 Nov 2026) · tag v1.10
Before. The cover's top line "THE PLAN · INITIATED BY UNCLE JAMES" (and the same for the Executive Summary, "COMPANION VOLUME A" to G and "THE SOURCES"); the companion volumes' series line "Saving Singapore's Hawker Heritage · a library of standalone documents"; the Executive Summary's and the Plan's cover title "Saving Singapore's Hawker Heritage"; Your Table's masthead line "SAVING SINGAPORE'S HAWKER HERITAGE · YOUR TABLE"; every page's running footer "SAVING SINGAPORE'S HAWKER HERITAGE · …"; every document's title "Saving Singapore's Hawker Heritage — …" in Document Control and the PDF metadata
After. Two series. The Executive Summary, the Plan and the Budget Book carry "SAVING OUR HAWKER HERITAGE · … · BY UNCLE JAMES · A SILICON WOK PROGRAMME", the cover title "Saving our Hawker Heritage" (the Executive Summary and the Plan) and the title "Saving our Hawker Heritage — …". Volumes A and C to G, Sources and Your Table carry "THE HAWKER LIBRARY · … · BY UNCLE JAMES · A SILICON WOK PROGRAMME" and the title "The Hawker Library — …". The companion volumes' series line is the series' own line: "The Hawker Library · The open reference on Singapore's hawker trade: its history, its prices, its incomes and its sources." (B: "Saving our Hawker Heritage · A citizen's plan, built on the Library. …"). Each running footer names the series: "THE HAWKER LIBRARY · FOLLOW THE PLATE", "SAVING OUR HAWKER HERITAGE · THE PLAN"
Why. Uncle James's decision of 2 Nov 2026: split the work into a Hawker Library, the evidence, which never argues, and a Saving our Hawker Heritage Plan, built on it. Only the connective cover and series text changes; the canonical text is not retitled
Before. §1.1 named one blank, the register of trades at risk, and gave the pay as measured once (MOM's $1,470 median for hawkers who work alone, 2021)
After. A second paragraph, "A second blank sits beside it: what hawkers earn": the Government does not track hawkers' earnings, because they are self-employed (MSE, Apr 2022); a fair reason that still leaves a gap we can close together; at the middle one stall's household takes home about $2,700–3,500 a month in 2026 prices, often for two people's work; the top 1% may keep $15,000–30,000, a plausible range; one median cannot describe a trade like that; a pointer to What hawkers earn and to §7.7
Why. The income findings, put as a gap to close together with NEA, not a failing (Uncle James, 2 Nov 2026)
Before. "Pay the goodwill." … "It is never priced on the location, which is the public's, and NEA may cap it. The veteran finally has a five-figure reason …"
After. Between those sentences: recipe sales show how wide the range runs, from $20,000 for a well-known Hokkien mee stall's stall and recipes (2020) to $500,000 to $1 million asked or offered for a few famous stalls' (2022–23); prices paid or asked, not measured incomes (grade C); they support a capped payment held in escrow, and give NEA a basis for where to set the cap
Why. Recipe prices give the cap on the goodwill payment a basis; no change to the ask
Before. The row "The spread, not only the median": "A median rent can hold for a decade while the hawkers paying it change. The spread shows who pays what, and settles the argument over fees with numbers (§2.2, §2.10)." The Plan's short form: "rent, fee and income spreads by centre and stall type (§2.10)"
After. The row adds: "For income, no series exists yet. The Library's graded estimate puts one stall's household at about $2,700–3,500 a month at the middle, and perhaps $15,000–30,000 at the top 1% (What hawkers earn). Its grades and its method are offered as a template NEA could adopt, and then improve on with the data it holds." The short form adds "with the Library's graded estimate of what hawkers earn offered as a template"
Why. The ask already exists; the new evidence strengthens it, and the grades and the method are offered to NEA as a starting template. No change to the ask, its instrument or its cost
Before. The old umbrella title, "Saving Singapore's Hawker Heritage", on the change log's cover and title ("Saving Singapore's Hawker Heritage — Version and Change Control Log") and in its opening note; Document Control's independence line "Hawker Heritage is an independent initiative stewarded by Silicon Wok". Outside the PDFs: the website's home heading and page title, the dataset name on the download page and the film's label, README.md, CITATION.cff, the simulator's eyebrow and note, the model's docstring and the canonical sources' titles
After. "Hawker Heritage, a Silicon Wok programme" (or "Hawker Heritage" where space is short): the change log is "Hawker Heritage — Version and Change Control Log", "one of the twelve documents"; the independence line reads "Hawker Heritage is an independent Silicon Wok programme", as on the website. The stall names stay: Stall 01 the Hawker Library, Stall 02 Saving our Hawker Heritage (the Plan). The website's home section is "Three stalls"
Why. The old umbrella title is retired; the collection is Hawker Heritage, a Silicon Wok programme, and the two disclaimers now read alike (Uncle James, 2 Nov 2026)
Before. Lower-case "the library" for all twelve documents: "Document 1 of the library", "Document 2 of the library", "document 1 of the library", "Volume G of the library"; "published as a small library with three doors" (ES.1); Document Control's row "The library" and its note "published beside the library"; "the references for the whole library", "for the library", "for every document in the library", "Sources, the library's list of references" (the sources notes of the Plan and of A to G, F.5); "where every lever in the library is printed" (the levers); "The library's text" (F.4); in S, "every reference the library cites", "in the library's order", "where in the library it is used", "Three marks run through the library", "found a link in the library", "any later document of the library"; the descriptions of G ("behind the library") and S ("Every reference the library cites")
After. "Document 1 of Hawker Heritage" and the same for 2 and G; "published as twelve documents with three doors"; the row "Hawker Heritage", "one of the twelve documents"; "for all twelve documents", "Sources, their list of references" or "the list of references for all twelve documents"; "every lever in the twelve documents"; "The documents' text"; in S, "every reference the twelve documents cite", "in Hawker Heritage's order", "where in the twelve documents it is used", "run through the twelve documents", "a link in Hawker Heritage", "any later document of Hawker Heritage"; "behind the twelve documents", "Every reference the twelve documents cite"
Why. Since 1.10 "the Library" is Stall 01; "the library" for all twelve documents now reads as that stall. Where it really means the Hawker Library it stays (Uncle James, 2 Nov 2026)
1.9.3 1 November 2026 3 changes · shared version
The time-bound body was the UNESCO nomination committee of 2018–20, not FMAS. FMAS is the trade's standing federation and carries on. The library now says exactly what MCCY was asked and answered on 7 Oct 2026, and speaks of NEA and its Hawker Centres Group as the partners whose work it builds on. No proposal, figure or claim changes.
Approved by Uncle James · tag v1.9.3 · release commit 9e7498e
Before. Two owners, one table (§3.8; A.5, B.14): the 14-member committee that won the inscription "lapsed" with it. The ask for a permanent committee (§7.7, B.7, ES.6, the Plan's short form, A.7): "the 2018–20 nomination committee of NHB, NEA and FMAS", "brought back" or "asked back as a permanent body". The A.5 opener: "the committee that won the inscription and then lapsed"
After. "co-chaired the 14-member UNESCO nomination committee that won the inscription. That committee was set up for the nomination alone, and the public record shows no meeting after the inscription. FMAS itself carries on, as it has since 1989." The ask names "the 2018–20 UNESCO nomination committee", co-chaired by NEA, NHB and FMAS, "brought back as a permanent body", and says FMAS itself is a standing federation. The A.5 opener: "the UNESCO nomination committee that won the inscription, set up for that task alone"
Why. Make plain that the time-bound body was the UNESCO nomination committee, not FMAS, which is a standing federation (Uncle James, 1 Nov 2026).
Before. §3.8 (A.5, B.14): asked "whether a formal mechanism should be set up", the Government "described the two agencies' work side by side" and, the line said, named no mechanism. §1.1 (A.4): "no count of practitioners or of dishes at risk (MCCY, 7 Oct 2026)". The website's card for the permanent committee: MCCY named no mechanism of a joint kind
After. "Asked in October 2026 whether a formal mechanism should identify the trades at risk, MCCY pointed to the schemes already running and left the question open (MCCY, 7 Oct 2026)." §1.1 (A.4): "no list of trades or practices at risk (MCCY, 7 Oct 2026)". The card: the same wording as §3.8
Why. Say exactly what MCCY was asked and answered on 7 Oct 2026 (Uncle James, 1 Nov 2026)
Before. §1.1 (A.4): "A scheme that reports enrolment is reporting its budget"; "Both are true; only one is reported"; "He may charge what he likes, as long as it is $3.50"; "The blank is on UNESCO's record too: … named the threats and gave no count". A.2.1: "a token stipend". A.4, D.14: "for what NEA could not do". B.7: "The number Parliament was told does not exist". ES.1: "The official reassurances … Four of them leave out the part that matters". Ideas credited to a party: the ballot "the Progress Singapore Party proposed in Parliament in 2024" (B.7), "credited to the PSP's 2024 proposal" (A.7), "Two Progress Singapore Party MPs" (ES.5, §1.3), and party labels on the Members credited in A.3, D.14, D.15 and F.9
After. "A count of outcomes would show what works"; "Both are true, and both are NEA's own findings"; "He sets his own prices, but the $3.50 value meal needs approval to change"; "named the threats but could not yet count them; the next one could"; "a small stipend"; "for something a price tender cannot do"; "The count Singapore does not yet keep"; "The official figures … Four of them need a second number beside them". "the ballot proposed in Parliament in 2024"; "credited to a proposal made in Parliament in 2024"; "Two Members, Leong Mun Wai and Hazel Poa"; the Members are named without their parties. The sources, Hansard and the parties' own published speeches among them, are unchanged
Why. A constructive tone towards NEA and its Hawker Centres Group, whose work the plan builds on; no change to any fact, figure or ask (Uncle James, 1 Nov 2026)
1.9.2 1 November 2026 2 changes · shared version
The pen name on every cover. 1.8 changed the byline, Document Control and the metadata, but the top line of every cover still gave my full name; it now reads Uncle James. No proposal, figure or claim changes.
Approved by Uncle James · tag v1.9.2 · release commit d28eda7
Before. The cover's top line: "THE EXECUTIVE SUMMARY · INITIATED BY", "THE PLAN · INITIATED BY", "COMPANION VOLUME A · INITIATED BY" (to G) and "THE SOURCES · INITIATED BY", each followed by the initiator's full name
After. "… · INITIATED BY UNCLE JAMES" on every cover
Why. the author's request: his pen name on every PDF. 1.8 changed the byline, Document Control and the PDF metadata, but not this line
Before. Two past rows still named him: the 1.3 row on §3.5's retention target ("for" his first name "to confirm") and the 1.7 row on "the plan" (the Plan's cover line, with his full name)
After. "TARGET CHANGE FOR UNCLE JAMES TO CONFIRM" and "THE PLAN · INITIATED BY UNCLE JAMES". No other past row changes, and the released PDFs' fingerprints are untouched
Why. the author's request: his pen name, in past rows too
1.9.1 1 November 2026 no change to this stall · shared version
The site's address and the contact in the papers are now Silicon Wok's. No proposal, figure or claim changes.
A shared version: no row touched Saving our Hawker Heritage's documents. They were rebuilt with the shared version number only, and their released fingerprints are recorded below.
Approved by Uncle James · tag v1.9.1 · release commit 9d75eca
1.9 1 November 2026 1 change · shared version
Hawker Heritage is now stewarded by Silicon Wok, the home of my writing as Uncle James. No proposal, figure or claim changes.
Approved by Uncle James · tag v1.9 · release commit 3c99397
Before. "Hawker Heritage is an independent initiative stewarded by Silicon Tribe."
After. "Hawker Heritage is an independent initiative stewarded by Silicon Wok."
Why. the author's decision: Silicon Wok is the initiative's home
1.8 31 October 2026 5 changes · shared version
A pen name and a shorter biography. From now on I write as Uncle James, and the library, the site and this log carry that name. The note from the author no longer names the places I have worked. No proposal, figure or claim changes.
Approved by Uncle James · tag v1.8 · release commit d6fcec9
Before. How to read this paper – a note from the author: his career named the government agency he worked at, the company he founded and the town of its factory; signed with his full name
After. "I began my career in the Singapore Government, working on industry development; I then spent years as an early-stage venture capitalist and angel investor; and I later founded a company that designed electric motorcycles in Singapore and built them, battery packs included, at its own factory in the region." Signed "– Uncle James, October 2026". The rest of the note is unchanged
Why. the author's request: pen name and an anonymised biography
Before. Initiated by, and the PDF author: the initiator's full name
After. Uncle James (versions.json people.initiator), in every document and in the change log
Why. the author's request: pen name and an anonymised biography
Before. About named him in full, with a bio naming his universities, the government agency, the venture firm and its co-founder, the companies he founded and helped build, the acquirer, his employer and a board seat; citations read "Surname, I."
After. "…a plan initiated by Uncle James…"; the bio reads: "Uncle James is based in Singapore. He trained as an engineer and studied public policy. He began his career in the Singapore Government, then spent years as an early-stage venture investor and a founder, most recently of an electric-motorcycle company whose technology and team were later acquired. He now works in the motor industry, and writes as Uncle James at siliconwok.com." The next paragraph is kept. Citations name the pen name as written: "Uncle James (2026)." The same in the build brief (website/BRIEF.md §7.8, §12) and the mock-up. The film's end card reads "Initiated by Uncle James" in its source; the rendered film is not rebuilt in this release
Why. the author's request: pen name and an anonymised biography
Before. Copyright, suggested credit, citation (surname and initial) and "initiated by" in the author's full name
After. "Copyright (c) 2026 Uncle James", a pen name; the licence terms are unchanged. CITATION.cff names the author by alias, "Uncle James"; README cites "Uncle James (2026)" at version 1.8
Why. the author's request: pen name and an anonymised biography
Before. approver and decided_by in the author's full name in every entry, and his first name in past rows' reasons; the 1.1 row on voice named the company whose updates were studied
After. Uncle James in every entry, past ones included; the 1.1 row reads "his blog, essays and company updates". No other past row changes, and the released PDFs' fingerprints are untouched
Why. the author's request: pen name and an anonymised biography
1.7 30 October 2026 27 changes · shared version
Plain words. Three readers found 841 places in the library where a diner, an MP's aide or a journalist would have to read twice, and 285 where they could not tell what was meant. I settled the nine numbers that did not add up first, then applied eight fixes by rule – one dash, one name for the plan, a name beside every id, a gloss at the first use of every term, the metaphors and the code out, long cells cut to lines, one sense per word – and rewrote the 285 by hand. No proposal, figure or claim changes.
Approved by Uncle James · tag v1.7 · release commit a99b0dc
Before. about half ($10.17M) to the heritage-worthy stalls; about a seventh ($2.64M) to every stall; about a third runs the institution
After. 57% ($11.25M, net of programme revenue) to the heritage-worthy stalls; 15% ($3.00M) to every stall at the pilot centres; the other 28% ($5.43M) runs the institution
Why. The three shares summed to 97% and "a third" contradicted B.1's 27.6% for the operating budget. Recomputed line by line from the funding model's ledger, with the 10% contingency; they now sum to the $19,683,808 total
Before. the corpus keeps its real value in 71 of 100 futures; in 20 some year runs short
After. ... in 20 some year runs short, though in 2 of those the corpus still keeps its value; and in the other 11 every year is paid but the corpus ends Year 20 below its real value
Why. 71 and 20 left 9 futures unexplained; the two overlap. The funding model now reports the joint outcomes (2.5% kept value with a short year, 10.7% lost value with none)
Before. A Relief-Hawker Corps to break the seven-day grind
After. A Relief-Hawker Corps to break the six-day grind
Why. The model's stall opens 26 days a month, 13 hours a day: the "13-hour, six-day week" of ES.2
Before. a fee to about 1.25 participants for every cooked-food stall, the 150 of the 120-stall centre
After. a $150 fee to about 1.25 participants for every cooked-food stall (150 participants at the 120-stall centre the ledgers were first drawn for)
Why. "The 150 of the 120-stall centre" ran the fee and the head count together: $150 a participant, 1.25 a stall, 150 people at 120 stalls (the ledger's rates)
Before. the site's About page, "eight companion volumes" (the covers said seven)
After. seven companion volumes, counted from versions.json everywhere (A to G); Sources and Your Table named beside them
Why. The count differed between the documents and the site. The covers already counted from the register; the About and Downloads pages now do too
Before. em dashes (—) in running text, about 900 of them
After. spaced en dashes ( – ), with an unspaced one for an empty cell; check_layers.py now fails on any em dash in reader text outside an official title that has one
Why. The house style is spaced en dashes. Plain-English audit, fix 1
Before. the framework, this framework; THE PLAN · A POLICY FRAMEWORK on the Plan's cover
After. the plan, this plan; THE PLAN · INITIATED BY UNCLE JAMES
Why. One name for one thing. "Framework" stays only where it names the 1.0 document and its versions. Fix 2
Before. a cross-reference printed its id alone ("see D.4.1")
After. the first time an id appears in a section of prose it carries its section's name, in italic ("D.4.1, How much rent can a stall bear?"); the bare id after. Not in tables, cards, contents, pointers or runs of ids
Why. Readers need the thing's name, not only its id. Fix 3, at build (src/layers/plain.py)
Before. abbreviations and policy terms (HDP, AMR, median, tranche, escrow, means-tested ...) used without a gloss
After. a plain gloss in brackets at the first use of each in each document, from the glossary (113 terms, 68 of them new); the site's pages carry the same
Why. Fix 4, at build (src/layers/plain.py; data/curated/glossary.json gains gloss and abbrev). The Executive Summary stays at eight pages
Before. addon_programme, centres_vetoed, four_centres, sample_stall(), plate_model.pnl(), centre_ledger(), run_handover(), PARAMS keys in F.3; ASSUMPTION in capitals in B.9
After. plain words (each centre's sizing; the model's profit and loss; the funding model's ledger ...); F.3 names each parameter; "assumption" in lower case in B.9 and the funding notes
Why. Code out of prose; the identifiers stay in G.1's register. Fix 8
Before. rails that stop one station short; the engine on the rail; plumbing and pressure; the incubation runway extended to touchdown; a bench of locum cooks; a rent cliff and a glide to band; trust deposits
After. plain words: each scheme stops short of the cook's income; the money the HSS lacks; a pool of relief cooks; the rent phased in over three years; gifts first. §2.3 is now "The Hawkers Succession Scheme, Extended – Pay the Veteran for the Goodwill". "Lighthouse" stays, explained at its first use as the first two pilot centres
Why. Plain-English audit, fix 5: a metaphor that needs the policy context to land. Uncle James, 30 Oct 2026, 06:43: the eight rule-based fixes approved, with the twenty rewrites of the audit as written
Before. table cells of 100 to 500 words, one run-on sentence each; D.7.4's note of about 480 words in one block
After. one fact or one action a line in every long cell; §6.1's Year 0 steps numbered in the order they happen; D.7.4's note a short list under "How to read the national table"; B.9's assumptions table prints one sentence a line
Why. Plain-English audit, fix 6: dense table cells become short lines
Before. "floor" for an income guarantee, a minimum payment, a wage and a stall's site; "tier" for the two tiers, for Bronze to Platinum and for centre Tiers A–C; Band and Tier for the same rent letters; goodwill as kindness; a goodwill sum "retired"
After. the floor is the Protection Floor only (an income top-up, a $600 minimum, the Progressive Wage, site costs); Bronze to Platinum are grades; centres are Class A–C and the rents Bands A–C; goodwill is money only; a goodwill sum is paid off; "a stall with long queues" for a strong business
Why. Plain-English audit, fix 7: one sense per word
Before. 19 sentences the audit marked High, among them the 110-word list of decisions in ES.6 and "the endowment draws 3.5% a year ... printed rather than hidden"
After. all 19 rewritten as the audit proposed, the ask as a numbered list of nine decisions; the Executive Summary stays at eight pages
Why. Plain-English audit: the High rows, in reader order
Before. 106 sentences the audit marked High (98 in the Plan's sections, 8 in the ledgers' row labels), among them the one-sentence ballot record of §1.3.2, the gate criteria of §6.5, the safeguards of §4.6 and the "stock and flow" of §7.8
After. each rewritten, the ballot record in three blocks (what was proposed, what the Government said, where I stand), the safeguards and the gate as numbered lists, every proposal card and the ask table in the same plain words; the curated proposal data quotes the new text
Why. Plain-English audit: the High rows, in reader order. The Plan ends at 46 pages, as with the glosses alone; Uncle James accepted 45 or 46 with the full glosses (30 Oct 2026)
Before. the volume openers' 15 High rows: "Government's rails come first", "the record is a decade of right rails without engines", "it survives a 2020", "the reserve must cover the bad tail"
After. plain bottom lines; the Budget Book's rise of $4,693,468 in four numbered reasons; CCCS, FSI, PMET and O&M spelled out
Why. Plain-English audit: the High rows, in reader order
Before. 30 High rows, among them "a contributing counterparty", "flows on", "little to annuitise", "the incidence rule", a 330-word note and a 150-word list of open questions
After. someone who pays a share of his CPF; spills over into retirement savings; C.6 "Enough to Retire On? Little to Turn into Monthly Income"; who ends up with the money; the note in four short paragraphs; the questions numbered; "bloc" becomes "group" throughout
Why. Plain-English audit: the High rows, in reader order
Before. 108 High rows in the model's prose and the levers, among them "custom", "recapture", "a lossy transfer", "a regression ... each driver scaled to one standard deviation", "a revolving reserve ... the P95 payout of the open book", a 400-word paragraph in D.5.4 and the funding model's "draw, close and bond ladder" undefined
After. customers; the landlord takes it back at renewal; the price rise loses $3,140 on the way, to customers who stop coming and tax; the factors and a typical-sized step; a reserve that is topped up, with P95 explained; D.5.4 in short paragraphs; B.9 defines its four terms before it uses them; the 5,000 futures as a list
Why. Plain-English audit: the High rows, in reader order. Every number is the model's, unchanged
Before. ASSUMPTION, ASSUMED, DERIVED and ILLUSTRATIVE in capitals in the model's notes and in D.4.2
After. in lower case, as in B.9; the site's assumption chips and the mock-up's read the tag in any case
Why. Fix 8, carried to the notes M7 left: one style for the tag
Before. 'rails' fifteen times, 'bench', 'engine', 'cliff' and 'levers'; 'tiers only go up' and 'centre tier'; supports run together in one 47-word sentence; the page's metadata said version 1.4
After. 'the CDC voucher system', 'schemes', 'a pool of cooks'; 'grades' for Bronze to Platinum and 'centre class' for bands; the six supports as six short sentences; sentences over 30 words split; the version is 1.7, and a test fails if it differs from versions.json
Why. Site-copy audit of 30 October 2026: the site's own words, in step with the papers' 1.7 names and plain style (day.json, day-view.ts and the page's intro)
Before. two 36-word opening sentences; '$500K'; 'the draw' and 'bond ladder' unglossed; 'the engine'; a claim that a till dollar 'does more than any single big cheque', which no figure backs
After. four short sentences; '$0.5M'; the bond ladder explained in brackets; 'the funding code'; the unbacked claim dropped
Why. Site-copy audit of 30 October 2026: the site's own words, in step with the papers' 1.7 names and plain style
Before. a 59-word sentence listing the library; 'Tell us' on a page that says he works alone; a 39-word sentence of three conditions ending 'Fail twice'
After. the library in five short sentences; 'Tell me'; the three conditions for a launch as three sentences, and what fails twice made clear
Why. Site-copy audit of 30 October 2026: the site's own words, in step with the papers' 1.7 names and plain style. The bio paragraph on /about/ was replaced earlier and is not touched
Before. 'All three papers' on three pages and 'the papers' on each proposal page; 'Follow the money: try it' on a button for the Follow the Gift page; a column headed 'A year'
After. 'The whole library' and 'the library'; 'Follow the Gift: try it'; 'Amounts a year'. Every page address is unchanged
Why. Site-copy audit of 30 October 2026: the site's own words, in step with the papers' 1.7 names and plain style. The library is eleven documents, not three papers
Before. centres in 'Tier A–C'; 'Lighthouse Centres' and 'Addendum A' and 'Addendum B' as the names of proposal and figure groups; 'rails', 'cliff' and 'second tier'; 'tier' for Bronze to Platinum; 'prep co-operative'; the Plate Tracker as 20 indicators with 10 in part
After. 'Class A–C' centres; 'Year Zero and the Lighthouses and Pilot Centres', 'Hawkers and Social Protection' and 'Follow the Plate' as group names; 'CDC voucher system'; 'grade' for Bronze to Platinum; the old heading of §2.3 replaced by its 1.7 heading; 21 indicators, 11 in part, 7 not at all. Proposal addresses are unchanged
Why. Site-copy audit of 30 October 2026: the site's own words, in step with the papers' 1.7 names and plain style. Strings that the papers quote word for word are not touched
Before. cover subtitles of 52, 50 and 46 words, with a list of seven inside a clause, five semicolons, and 'levers' and 'downside mode' as shorthand
After. the same facts as short sentences; 'the options for the Hawker Protection Floor' for 'the levers'; the register's own description (versions.json 'about') in five sentences
Why. Site-copy audit of 30 October 2026: the site's own words, in step with the papers' 1.7 names and plain style
Before. version 1.2, 'a policy framework', nine PDFs, six companion volumes, $14.99M and 65%, the 'Heritage Watchlist', the page counts of 1.2, no row for G, Sources or Your Table
After. version 1.7 of 2026-10-30, a plan in twelve PDFs, seven companion volumes, $19,683,808 and 72.4%, the Regulars' List, the page counts of 1.7 and a row for G, Sources and Your Table; the 1.5 entry's longest bullets cut into lines. check_publications.py now fails if README.md or CITATION.cff lacks the register's version
Why. Site-copy audit of 30 October 2026: the site's own words, in step with the papers' 1.7 names and plain style. The two files had stayed at 1.2
Before. the Plan's colophon said "Six companion volumes" (there are seven); the Part VI divider read "Year Zero & the Lighthouse Centres" (§6.3 says "The Lighthouses and the Pilot Centres"); the model said a mall food court is "outside CDC rails" and the simulator said Makan Vouchers reach stalls "on CDC rails"; AGENTS.md described a library of nine PDFs with six companion volumes
After. seven companion volumes, and check_layers.py now fails if any printed count differs from the volumes in versions.json; the divider reads "Year Zero, the Lighthouses & the Pilot Centres", and the proposal group in the data reads "Part VI · Year Zero, the Lighthouses and the Pilot Centres"; the model says "outside the CDC voucher system" and the simulator "in the CDC voucher system" (plate_model.py and model.js changed together, parity unchanged); AGENTS.md describes twelve PDFs and seven companion volumes
Why. The last four drift fixes before release, found in a final read of the library against versions.json, the section titles and the site's own words. No proposal or figure changes'
1.6 30 October 2026 9 changes · shared version
Four calls 1.5 left open, settled on the evidence. NEA already staggers the rent when a subsidised stall is assigned, so the HSS ask narrows to writing that stagger into the Scheme's own terms, and Year 0 asks whether it already applies. The 2012 Panel reported in April 2012. MOM's own rules answer the foreign-worker question, so it is now a recorded fact, not an open one. And Liang Eng Hwa's 2018 question on hawkers managing their own centres gets its credit. The model does not change.
Approved by Uncle James · tag v1.6 · release commit 40c0a28
Before. NEA staggers the rise to assessed market rent "on the assignments it allows, but not under the HSS"; the successor's rent "can rise several-fold overnight"
After. NEA phases the rise in over three years when such a stall is assigned (§2.3); hand-over outside the family is still a rent cliff, even if a three-year slope
Why. NEA's assignment form (11 Dec 2020) and family-transfer form (28 Aug 2018) both stagger the rise over three years; no NEA source says the HSS denies it. Uncle James, 30 Oct 2026, 01:21: the four open calls, settled as the research note of 30 Oct 2026 recommends
Before. Give the HSS successor the stagger NEA gives other assignees; under the HSS the successor pays assessed market rent "from the start"
After. NEA already glides the rent where the jump is steepest; the ask narrows to NEA writing its own three-year stagger into the Scheme's terms, to the band at a pilot centre. Year 0 asks whether it already does; if so, the line becomes a credit, not an ask
Why. The HSS FAQs (14 Mar 2022) say only that the successor "will pay assessed market rent to NEA"; "from the start" was ours. For non-subsidised veterans NEA's renewal moves rent towards AMR by no more than $300 a term (4 Nov 2024). Uncle James, 30 Oct 2026, 01:21: the four open calls, settled as the research note of 30 Oct 2026 recommends
Before. The HSS successor "jumps straight" (§1.2) or "jumps immediately" (§2.3) to assessed market rent
After. The successor pays assessed market rent
Why. Follows from the stagger on assignment; NEA is silent on whether HSS hand-overs use it
Before. The §2.3 card and the ask tables (§7.7, B.7): "the three-year rent glide NEA already gives other assignees", "rent gliding to band over 3 years"; Year 0 asks whether the Scheme's assignments "can already run under NEA's ordinary three-year stagger"
After. NEA's own three-year rent stagger, written into the Scheme's terms; Year 0 asks whether hand-overs already get it, and if so the line is a credit, not an ask
Why. The card and the ask tables follow §2.3
Before. The Hawker Centres Public Consultation Panel (2011–12)
After. The Panel (November 2011 to April 2012); the Ministry accepted most of its report on 10 May 2012 (The Straits Times, 11 May 2012, as reported)
Why. The Hawker Centre 3.0 Committee's report (Feb 2017, p. 9) and The Straits Times (11 May 2012) both put the final report in April 2012; the Panel's last meeting was on 27 Mar 2012, so February cannot be the final report
Before. Panel report: "chaired by Elim Chew; 18 members; met Nov 2011 – Mar 2012", ¶12–19. The assignment form without a web address
After. The Panel's full citation line (formed Nov 2011; seven meetings to 27 Mar 2012; undated; submitted April 2012, per HC 3.0, p. 9), ¶2 and ¶12–19; The Straits Times, 11 May 2012, added; HC 3.0 p. 9 added; the assignment form's web address and NEA's family-transfer form (28 Aug 2018) added
Why. The sources for the calls above, each in Sources once
Before. The State built hawker centres from the early 1970s "to move street hawkers indoors, for hygiene and food safety, and resettled the last of them by 1986" (uncited)
After. To resettle street hawkers, "so as to improve hygiene and food safety standards" (Masagos Zulkifli, Official Report, 19 Nov 2018), finishing by the mid-1980s (Panel report, ¶2)
Why. A primary source, not a party publication; the Panel says 1985 and others 1986, so "by the mid-1980s"
Before. The social-enterprise row named the 2018 questioners but not the alternative put to the House
After. Liang Eng Hwa (PAP) asked on 19 Nov 2018 whether hawkers could "come together to self-manage" a centre on a strata-title model; Dr Amy Khor replied the Government was "open to any model" that met its social objectives (Official Report, 19 Nov 2018, Vol 94, No 85)
Why. Credit where it is due; dated to 2018, not the 2024 debate, where Liang did not raise it
Before. A chain's stall may draw on its operator's foreign-worker quota, up to 35% of staff (Chua, IPS, 2024; "to be confirmed with MOM"); a question, not a proposal
After. MOM's rules answer it: no work permit holders at any hawker centre, operator-run included (MSE, 21 Apr 2023), LTVP holders since January 2025; in coffeeshops and food courts only the food-shop licensee may hire foreign workers, for its own stalls, at one to two locals (MOM FAQ, 14 Mar 2024; MOM, 25 Feb 2026). Recorded as a fact, not an ask
Why. The question is answered; the Government kept its rules on 10 Sep 2026, so a new ask would repeat it
1.5 29 October 2026 36 changes · shared version
Three doors, one library. A politician reads the Executive Summary and the Plan, which falls from 72 pages to 44: every proposal now opens with the same card (what it is, who it is for, what it costs, who pays, what it asks of the State, what Year 0 must confirm and where it is proved), and every block the Plan no longer prints is printed in full in a volume, with its old route redirected. An official reads volumes A to G: A becomes The Record (who helps, what was said and done since 2012, the objections answered), and a new volume G, The Technical Paper, prints every dataset, simulation and sensitivity straight from the models, with the downloads and a parity statement. A diner reads Your Table, four pages on the plate, the cook and what a regular can do. And every reference the library cites is now printed once, in a document of its own, Sources: numbered, with where each is used and its web address, so that anyone can check us. The paper colour now runs to the edge of every page. No proposal, figure or claim changes.
Approved by Uncle James · tag v1.5 · release commit 511d6cf
Before. Each proposal opened with its chips and, for some, an 'In one minute' box or a 'Going deeper' line
After. Every proposal opens with a fixed card: what it is, who it is for, what it costs (Years 0–3, from key-figures or the funding model's ledger lines), who pays, what it asks of the State, what Year 0 must confirm, and where it is proved
Why. Uncle James, 29 Oct 2026: keep the Plan readable for politicians; the card says where every argument and number is proved
Before. An 'In one minute' box at the head of Parts II–VII
After. Retired; the cards and each part band's lede replace them (minute.json keeps the texts)
Why. The cards say the same thing per proposal; the Plan from 72 to 44 pages
Before. In full, 1,633 words
After. The lead, the five numbers, the fallacies box and the blank; 'This is not a training deficit' and 'The pay has been measured once' moved to the Record, A.4; 'Two pressures' (the RTS Link) to Follow the Plate, D.15
Why. Brief: cut §1.1's prose around the fallacies box first
Before. §1.2 in full, with the scheme-by-scheme table
After. §1.2 keeps its intro, the dated record with the panels credited, and the pattern; the full section, table included, is the Record, A.2.1
Why. Destination map: the lessons scheme by scheme go to A
Before. §1.3's lead, the Government's rails table, the intro of §1.3.1, the map and its read-across, the closing of §1.3.2
After. The map (Fig. ES.1), its read-across and a pointer; §1.3 in full is the Record, A.1.1
Why. Destination map: §1.3 to A in full; the Plan keeps the infographic and the half-page read-across
Before. In full
After. Lead, table and closing line; the box 'The one route to ownership has closed' moved to the Record, A.2
Why. Brief: §2.1 to a page
Before. In full, 3,106 words
After. Card, the five band rules and 'Why we think this works'; §2.2 in full is Follow the Plate, D.14 (rents, the newcomer's path, the operator-run premium, the unpublished fees, the fee subsidy); the Panel paragraph and the objections also in the Record, A.3–A.4
Why. Destination map
Before. In full
After. Card only; §2.3 in full is Succession Economics, E.3
Why. Destination map: §2.3 to E
Before. In full
After. Card, lead and At your table; §2.4 in full is Hawkers and Social Protection, C.12
Why. To bring the Plan to 44 pages; C already holds 'whoever the landlord' (C.7)
Before. In full
After. Card, the price-freedom and frame items, the blockquote and At your table; §2.5 in full (incidence, the tranche arithmetic) is Follow the Plate, D.15
Why. Destination map: §2.5 to D
Before. In full
After. Card and the opening paragraph; §2.6 in full is Follow the Plate, D.18
Why. To bring the Plan to 44 pages
Before. In full
After. Card, four of the bench items and At your table; the booking economics and the named deputy are the Budget Book, B.12 'The Relief Cook's Day, Costed'; the lead, the relief notice and the farm models are Hawkers and Social Protection, C.11 'When Illness Closes the Stall'
Why. Destination map: §2.7 to B (costing, the day rate) and C (medical leave)
Before. In full
After. Its first and last paragraphs; the three reasons also printed in Method, Levers and Data, F.9
Why. To bring the Plan to 44 pages
Before. In full
After. Card, its first paragraph and the two covers; §2.9 in full is Hawkers and Social Protection, C.10
Why. Destination map: §2.9 to C
Before. In full
After. Card, its first paragraph and the box 'What we count, and why not the median'; §2.10 in full is Method, Levers and Data, F.6
Why. Destination map: §2.10 to F, the metrics box kept in the Plan
Before. §2.11.1–§2.11.2 (Addendum A's A.7 and A.9) in full, with an 'In one minute' box
After. Card, the lead, the proportionality test, the proposals table and 'What hawkers should not get'; A.9's lead and what stays unasked. Both in full are a new section of Hawkers and Social Protection, C.9 (C.9.1, C.9.2)
Why. The floor's long argument belongs with the comparison it rests on; the Plan from 72 to 44 pages
Before. In full
After. Card, three of the circle items, a credit line to the 2012 Panel and the Hawker Centre 3.0 Committee, and At your table; §2.12 in full is Follow the Plate, D.16
Why. Destination map: the bulk-buying arithmetic to D, the Panel/HC 3.0 credit kept
Before. In full
After. §3.2: card, its record and the locum test ('Not the first record' to the Record, A.6). §3.3: card only, in full in the Budget Book, B.15. §3.4: card and its table, in full in the Budget Book, B.16
Why. To bring the Plan to 44 pages
Before. In full
After. Its intro and a short list of the targets; §3.5 in full is Method, Levers and Data, F.7
Why. Destination map: §3.5 to F
Before. In full, 2,076 words
After. Card, 'a compliment' and 'one way on, and only up'; the tiers, thresholds, what a listed stall gets and the cost are Follow the Plate, D.17 'The Regulars' List, Tier by Tier'; the intro, the at-risk count, why UNESCO needs it and the Registry are the Record, A.5
Why. Destination map: §3.6 to A (the UNESCO record, the count's purpose) and D (the tiers)
Before. In full, 1,441 words
After. Card, its first paragraph, the casting vote and the sunset clause (about 350 words); §3.8 in full is the Budget Book, B.14; the associations and 'Two owners, one table' also in the Record, A.5
Why. Destination map: §3.8 trimmed, the committee history and 'two owners, one table' to A
Before. In full
After. Card, its lead, the Fellowship's terms and two items; §4.2 in full (the stipend's costing) is the Budget Book, B.13; 'Set against what exists' also in the Record, A.6
Why. Destination map: §4.2 to B
Before. In full
After. Cards and the first items of each; §4.3 in full is Succession Economics, E.4 and §4.4 E.5, with §5.5's 'One charge per sale' and two of its items
Why. Destination map: §4.3 and §4.4 to E
Before. In full
After. §5.2: card added, text in full. §5.5: card, lead, three items and At your table; the rest in the Record (A.5, A.6) and Succession Economics (E.5)
Why. Destination map: §5.5's 15.COM 8.B.6 controls to A
Before. In full, 1,759 words
After. A short form naming the three centres and why, and a pointer; the ranking, the conditions, the centres' table and the case for more than one centre are Method, Levers and Data, F.8 'How the Pilot Centres Are Chosen'
Why. Destination map: §6.3 to F, the Plan keeps the three named centres and why
Before. §7.1 with its four-year paragraph, the bars and three tables; §7.6 with its endowment box
After. §7.1: its first paragraph and the totals table; §7.6: a short verdict (the draw pays part of each year, not most) and the one funding table; both in full in the Budget Book, B.1 and B.6
Why. Destination map
Before. A short paragraph of the asks, 'Who the money reaches' and 'And what we do not ask'
After. One table: the ask and its Budget cost, of whom, by when, and what we don't ask; the reasoning, by row, in the Budget Book, B.7
Why. Destination map: §7.7 becomes a table
Before. §7.8 downside and sunset in short; §7.9 in full
After. §7.8: the downside paragraph and a pointer (the sunset in the Budget Book, B.8); §7.9: its first line, a one-paragraph list and a pointer; in full as an appendix, Method, Levers and Data, F.9
Why. Destination map: §7.8 in short, §7.9 to F as an appendix
Before. The volumes printed their own sections and levers
After. Each volume ends with a part 'From the Plan, in full': the Plan's sections it owns, word for word, numbered on from its last id so that no existing id moves; a reference to one of them from another volume goes there
Why. One Plan, not two: the mechanism of 1.2 (sections cut from the Plan and placed in the volumes) extended
Before. Each opener's 'In this volume' and 'Where it comes from'
After. Updated for the sections each volume now prints from the Plan
Why. The volumes' new contents
Before. A lever's label named its Plan section through the volume's own rewrite (e.g. 'B.8', 'D.14')
After. The label always names the Plan's section ('the Plan, §2.2')
Why. Since 1.5 most Plan sections have a full copy in a volume, which the label is not about
Before. ES.1 A Crowded Centre, an Empty Register · ES.2 The Crisis in Five Numbers · ES.3 The Thinnest Floor · ES.4 Who Already Helps (the map) · ES.5 Three Moves · ES.6 What Each of Us Gets · ES.7 What It Costs, and What We Ask · ES.8 Where to Go Next
After. ES.1 What the Reassurances Leave Out (four, each with its source) · ES.2 Three Facts ($1,470 against the riders' $1,800; $372 a closed day; $18,900 above valuation; the health sentences kept) · ES.3 Two Tiers, Stated Once · ES.4 Three Moves, as three cards · ES.5 What It Costs, and Who Pays (the totals, 72.4%, the endowment verdict, the map) · ES.6 The Ask, and 30 June 2027 (with the routes table). Eight pages, as before
Why. Uncle James, 29 Oct 2026: the summary for politicians; the routes of 1.1–1.4's ES sections redirect to the section that now carries their content
Before. Contents with sub-lines for Parts I and II; the Plan's cover line
After. Contents without the two sub-lines, a shorter start note on the cards; volume list read from versions.json (G and Your Table appear when they are listed); the cover names the cards
Why. To keep the contents on one page
Before. Every PDF printed the cream paper colour only inside the page's margin box, a tinted block inside a white border
After. Every document: the paper colour now runs to the page edge (@page background in the shared stylesheet; the body transparent); cover and divider pages unchanged
Why. Uncle James, 29 Oct 2026: the white border was ugly
Before. The Executive Summary's cover: six companion volumes hold who already helps, the budget, the evidence, the levers and the method. ES.6's routes table sent a policy officer to volumes A to F; the Plan's cover listed the record, the money, the evidence, the levers and the method
After. Both covers count the volumes from versions.json (seven) and add the data; ES.6's routes table sends a policy officer to volumes A to G, with the Technical Paper for the data and the simulations (G)
Why. Volume G joins the library in 1.5, and the two documents a politician reads first should say so
Before. Each document's sources note: "The references for the whole library are collected in Method, Levers and Data (F.5)"; F.4's layers timeline gave "the sources of D.5.4 in F.5"
After. Each names Sources instead: the Plan's note, "the references for the whole library are in Sources … (S.1–S.238 for those the Plan cites first)"; C and D say their lists are in short with the S-numbers; F.4 says "listed in Sources"
Why. F.5 no longer holds the lists
1.4 29 October 2026 50 changes · shared version
The library now reads the Government's own record first (the 2012 Hawker Centres Public Consultation Panel, HC 3.0, the Official Reports of 12 March 2013 and 19 November 2018, Singapore's UNESCO periodic report, the ACPW report and Chua's IPS paper) and answers the objections it raises: hawker pay has been measured once, about $1,470 a month in 2021, and is reconciled with the model; the health evidence enters C and §2.9, where a day shut for illness costs a typical tendered stall about $372; ten reassurances are set against what they leave out, the count is kept as distributions, not medians, and one standing NEA–NHB–FMAS body is asked to own it. The Regulars' List counts orders, not plates, a Year 0 stall baseline sets its thresholds, and the four-year budget at three centres is $19,683,808. Granted, the record also cuts the plan down to size: on Chua's evidence a third of Heritage Plate licensees are assumed to fail each year, and the Government's match now covers part of the base (40%), not most of it.
Approved by Uncle James · tag v1.4 · release commit be677f3
Before. A regular is one named diner who eats at the stall at least weekly; Silver, Gold and Platinum at 50, 100 and 250 endorsements or 8%, 15% and 40% of the stall's weekly plates, at 1.5 visits a week
After. A regular is one named person who orders from the stall at least weekly; the share route counts orders, not plates: 10%, 20% and 50% of the stall's weekly orders, at 1.5 orders a week each; the counts route (20, 50, 100, 250 names) unchanged
Why. Uncle James, 29 October 2026: a family's takeaway is one order of several plates, so it lifts the plate count without lifting the count of regulars. The shares are set just below what 50, 100 and 250 regulars order at the typical tendered stall (10.4%, 20.8%, 52.1% of 720 orders a week), so it reaches each tier at about the same number of regulars as before
Before. (no plates an order; one plate a visit)
After. Plates an order by dish type, ASSUMPTION until the baseline: about 1.2 for drinks and snacks, 1.3 for noodles and rice, 2.0 for family dishes (kway chap, zi char, roast meats); dollars an order = the plate's price × plates an order; orders a week derived from plates a week
Why. No published Singapore figure gives plates or items an order at hawker stalls; the nearest, Hawkers Go Digital's $45 million in 5.4 million e-payment transactions in August 2023 (Enterprise Singapore, 26 Oct 2023), is about $8.33 a transaction, about 1.9 of the model's $4.50 plates, a ceiling since small orders are more often paid in cash
Before. (no stall-level baseline; the evaluator's baselines only)
After. A stall baseline in each pilot centre's year before it opens: a two-week till count at every consenting stall (orders a day, plates an order, dollars an order, the share of orders from regulars the hawker names), by dish type, gathered by the listening teams through oral interviews and a tally sheet, published only by dish type and centre; it sets plates an order and the share thresholds by dish type before the first listing, then gives the Plate Ledger its first stall-level numbers
Why. Uncle James, 29 October 2026: collect baseline data from oral interviews to re-establish plate counts by stall and dish type and dollars an order
Before. Year 0: A $404,800, net $834,955, 51.5% reaching hawkers, $33,700 a month; Year 1: A $1,451,285, net $4,598,196, 68.4%, $120,900 a month
After. Year 0: A $476,443 (the stall baseline at both Lighthouses, $65,130 before contingency: 867 interviewer hours at $50, tally sheets, a data assistant for 4 months), net $906,598, 47.4%, $39,700 a month; Year 1: A $1,482,338 (Tekka's baseline, $28,230), net $4,629,249, 68.0%, $123,500 a month
Why. Costed in funding_model.py (stall_baseline): 3 interviewer hours a stall at the listening sessions' rate ($150 a session, taken as 3 hours: $50 an hour), $20 of tally sheets a stall, a data assistant at the research assistants' $4,000 a month for about every 80 stalls; at every cooked-food stall as a ceiling; in part A, since it reaches no hawker as money
Before. $19,581,112 over four years at three centres; 72.8% reaching hawkers; $5,327,410 operating, about $111,000 a month; $4,590,772 (30.6%) more than 1.2
After. $19,683,808; 72.4%; $5,430,106, about $113,000 a month; $4,693,468 (31.3%) more than 1.2
Why. The stall baseline adds $102,696 with contingency: $71,643 in Year 0 and $31,053 in Year 1
Before. The two Lighthouses alone $16,152,899; Old Airport Road $5,542,161; Maxwell $4,384,879; four centres $25,123,273; five $29,508,152; a further centre $3,608,504
After. $16,224,542; $5,586,777; $4,412,940; $25,270,585; $29,683,525; $3,636,004
Why. Each centre adds its own stall baseline in the year before it opens
Before. Year 0 foundation grants $641,955; Year 1 bridge grants $3,578,196
After. $713,598; $3,609,249
Why. Year 0 is met by gifts for spending and foundation grants; Tekka's baseline falls in Year 1
Before. (no note on the baseline's or plates-an-order assumptions)
After. The stall baseline's interviewer hours and rate, plates an order by dish type and the order rate are named as our assumptions, with Hawkers Go Digital's totals as a ceiling only
Why. Every unverifiable claim is marked in its volume's sources note
Before. Heritage Plate licence income with no failures: about $90K in Year 3 and $3.0M a year from Year 7
After. A third of licensed outlets and packaged lines assumed to fail and pay nothing in any year (ASSUMPTION; tested at none and half): about $60K in Year 3 and $2.0M a year from Year 7; E.1.1 explains why the royalty is on takings
Why. Nigel Chua (IPS, June 2024, pp. 31–48): most licensed or acquired hawker expansions shrank or closed, and the one profit-share licence, Lao Chaozhou's, paid the founders nothing
Before. Base match covered 54% over Years 4–10 ($11.2M short; with HDB the stretch $28.9M); corpus in Year 20 129% (105% with half earmarked); 76% chance the corpus keeps its real value; 'most of the base'
After. 40% ($14.7M short; $32.4M); 121% (101%); 71%; 'part of the base, not the stretch'; at a 20% maintenance rate $5.8M short over Years 4–20 and the corpus at 87%; the §7.1 ledger for Years 0–3 unchanged
Why. Follows from the licence failure rate; the Seed Bank's 20% of licence income joins the continuing gifts
Before. Key figures: Old Airport Road $1,937,742 and $3,136,919 in Years 2 and 3; Maxwell $1,562,075 and $2,444,569
After. $2,026,842 and $3,285,419; $1,636,325 and $2,563,369, as §7.1's table; both totals now checked against the ledger
Why. data/curated/key-figures.json had not followed the 1.4 stall baseline
Before. Repairs and redecoration every six to eight years
After. Every six to eight years, as NEA now says, against five to seven in the Minister's answer of 19 Nov 2018
Why. Official Report, 19 Nov 2018: the answer gave five to seven; NEA's current page is newer
Before. Inscription "arrived just as the trade began to age faster than it could renew itself"; ES: "ageing faster than anyone is replacing them"
After. Masagos Zulkifli, 19 Nov 2018: "more than one third of our hawkers and their assistants will retire in the next 10 years"; that decade ends in 2028; median about 60 since 2018
Why. The median age was about 60 in 2018, 2021 and 2024, so 'ageing faster' was not shown; the Official Report gives the dated warning (Vol 94, No 85, p. 15)
Before. "No data" / "No official source measures hawkers' earnings"; C.5 Income: "No official figure"
After. About $1,470 a month in 2021, median gross profit of full-time own-account hawkers (MOM CLFS 2021, ACPW report Fig. 1.1, read from a chart), below P20 ($2,200), riders ($1,800) and private-hire drivers ($2,000); caveats: 2021, own-account only, before tax
Why. Correction 2: MOM has measured hawker income once, in the Advisory Committee on Platform Workers' report (17 Nov 2022)
Before. (no answer to the Hawker Centre 3.0 Committee's finding that hawkers did not flag pay)
After. Paragraph "The pay has been measured once": HC 3.0 (2017, p. 20) quoted fairly, answered by survivorship and the $1,470 chart; Singapore told UNESCO NEA ensures "decent livelihoods for hawkers" (Dec 2024)
Why. Addition B: meet the Government's own record rather than step round it
Before. (none)
After. §1.1: a box "What the reassurances leave out": ten claims, each with what it leaves out, the evidence and a one-line answer, sourced; ES.1: four of them (1, 3, 5, 9) as single sentences with four footnotes
Why. Addition G, Uncle James's fallacies table of 29 Oct 2026; the answers address the claim, never the person
Before. ES.2 two paragraphs of 9 lines; ES.3 closing paragraph; ES.4 a long paragraph naming Louis Chua and the repairs cycle
After. ES.2 and ES.4 shortened, ES.4 crediting the Government's own panels of 2012 and 2017 first; ES.3 adds the three health sentences (78 hours against the WHO/ILO 55; two in three restaurant workers in pain; $372 a closed day)
Why. Keeps the Executive Summary at 8 pages with additions G and K (checked by composing in a scratch copy)
Before. §1.2 opened on the table; the "Not new, and said so" list began with the PSP's ballot
After. A dated history: resettlement from the early 1970s; the 2011 restart ("a ballast"); the 2011–12 Public Consultation Panel (Elim Chew; ¶12–19) and the Hawker Centre 3.0 Committee (2016–17; eleven recommendations accepted 8 Mar 2017; Rec 5, Rec 9; no hawker on it; it backed operator-run centres); A.3 leads with both panels
Why. Addition A: credit the Government's own panels before any party
Before. Master Hawker Programme lesson: training without the recipe; SEHC row: "Deep, durable distrust"
After. HC 3.0's own reason added (set-up costs, lack of experience, p. 22); SEHC row: plough-back intent (CNA/TODAY, 20 Oct 2018), HC 3.0's backing, 2 → 7 of 114 → 16 of 123 centres, the 2018 questions from both sides, the Nov 2018 caps, and the Government's case: median age 43 at the seven SEHCs (2018)
Why. Additions A, I; even-handedness (U, V findings)
Before. FMAS "co-chaired the UNESCO nomination"
After. The 14-member nomination committee of 2018–20 was co-chaired by the CEOs of NHB and NEA and FMAS's President
Why. Correction 9 (Grace Fu, PQ reply, 15 Jan 2019; Malay Mail, 23 Oct 2018)
Before. NHB's inventory "does not say which stalls or dishes are at risk"
After. NHB reviews its 106 elements once a year in phases; ICH-050 last revised March 2019; the Registry offered as research for its next revision
Why. Addition D; V corrects Q's 'promise broken' framing
Before. "The subsidy is personal: it does not pass to a successor"; subsidised share "from about 40% to about 30%"
After. It "can be transferred to the immediate family members of the pioneer hawkers at the same low rent" (Koh, 13 Nov 2024); 53% (2013) → about 40% (2020) → about 30% (2023); the cliff is for a successor outside the family
Why. Corrections 1 and 4 (Official Report, 12 Mar 2013, Vol 90, Issue 13)
Before. Sale-scheme box: about 2,000 stalls; at expiry subsidised stallholders went back on subsidy
After. About 2,000 stalls in 15 centres, about 600 of them cooked food (MTI 2015); Grace Fu's options at expiry: back to subsidised rent, hand back for an ex-gratia payment, or assign at market rent
Why. Correction 4
Before. Four objections answered; the bands credited to the 2024 debate
After. The 2012 Panel's "pre-determined" rents by food type (¶17–18) credited first; a fifth objection ("subject to potential audit issues", Masagos 2018) answered by the operator-run centres' tastings and the published gate; the "business is poor" principle accepted and met
Why. Additions A and B
Before. (the newcomer's bill shown only as medians)
After. "What a newcomer actually pays": $1,800 bid against $1,450 valuation under the Nov 2024 rule, $18,900 above valuation over six years ($29,700 against $1,250); $8.01, $8.54, $9.08 an hour against the assistant's $11.64; KF Seetoh stated fairly
Why. Addition I; A4's newcomer_rent_path
Before. Conservancy and table cleaning: "about half as much as the median rent"
After. Rent, conservancy and table cleaning = 1.5× rent at comparable centres, 1.6× at operator-run ones; dishwashing, POS and other fees not published; 'double the rent' needs $650–730 a month more; the SEHC premium about 17% in 2018 ($2,000/$1,700) and 2025 ($1,700/$1,450), with the caveat
Why. Addition I (TOC, 15 Sep 2026; Official Report, 19 Nov 2018; our arithmetic)
Before. Glide the rent: "mirroring the glide Government already applies, in the other direction"
After. Give HSS successors the three-year stagger NEA already applies on assignments (assignment form, 11 Dec 2020; LegCo FS10/13-14); HSS FAQ: successor "will pay assessed market rent"
Why. Correction 7
Before. Paying for the cheap plate credited from 2021 (WP) and 2024 (PSP, WP)
After. Song Seng Wun, Walter Theseira and Donald Low credited first (CNA/TODAY Big Read, 20 Oct 2018)
Why. Addition J
Before. (no State precedent for master–apprentice transmission)
After. NHB's Heritage Apprenticeship pilot (Aug 2024, six months, three Malay crafts); HC 3.0 Rec 1 ("traditional hawker fare"); veterans' fear of "recipe leaks" (p. 14)
Why. Additions A and E
Before. §2.9 had no health evidence; C.5's "The body is the business" one paragraph
After. §2.9 "Why cover comes first"; C.5 "What the work does to the body" and "Why cover is a heritage measure": WHO/ILO long hours, Makki 2026, Tüchsen 2000, NUS stall fumes, IARC 2A, Leong 2024, Tang 2010, two dated closures; $372 a closed day ($342, $441), six weeks $13,400 (35%), a rent waiver $48 a day, a Corps cook $40; only the Corps reaches the commonest harm
Why. Addition K (H2's health findings; A4's illness_cost)
Before. Repairs and redecoration "every six to eight years"; HUP 2.0 consultations
After. Six to eight years (NEA, 2026; five to seven when Parliament was told, 19 Nov 2018); the Hawker Council names stall exhaust and fume capture as health items in HUP 2.0
Why. Correction 8; H2 note 3
Before. (the Ledger did not say why it avoids medians); §3.5 had five metrics
After. Box "What we count, and why not the median": distributions, flows and health measures by stall and centre type; §3.5 adds Hawker health and Flows rows and the incubation baseline (37 of 99 by June 2025; 8 of 34 in 2021)
Why. Additions H and J (metrics.md)
Before. Buying together "has not taken off"
After. The 2012 Panel's "central bulk purchases" (¶12) and HC 3.0 Rec 5 (2017) credited
Why. Addition A
Before. Masters choose what is recorded
After. NHB's own contracts vest "foreground IP rights" in practitioners, with two-step consent (periodic report, Dec 2024); Not for Sale (2013) credited
Why. Addition D
Before. The at-risk count "sits beside the Registry's list in Singapore's reports to UNESCO"
After. "Why UNESCO needs the count": the form asks for practitioner demographics and threat severity; the 2024 report gave no count and promised better monitoring of bearers and practitioners; the count offered for the 30 June 2027 report; HC 3.0 Rec 9 credited; the report's "more than half … second and third generation" set against the heritage proxy
Why. Addition D
Before. Government director "nominated by NEA or MSE, as they choose"; NHB liaison without a vote
After. Nominated jointly by MSE and MCCY; the NHB liaison may table the UNESCO items; "Two owners, one table": no joint mechanism (MCCY, 7 Oct 2026), no hawker-culture KPI, the 2018–20 committee lapsed
Why. Addition F (G2's governance findings)
Before. Kay Lee's "move to a central kitchen"; a subsidised hand-over never priced on a subsidy "the successor will not inherit"
After. The central kitchen and ten outlets never happened and Kay Lee ended as one shop (Chua, IPS, 2024); a child keeps the pioneer rent and needs no mortgage; Balakrishnan's "the key principle is personal operation of the stall" (2013)
Why. Corrections 1 and 6; addition J
Before. "one founder's sons can run rival stalls"
After. A founder's son and grandson; the 2008 suit ended with $1,000 nominal damages; the royalty on takings because Lao Chaozhou's profit-share licence paid nothing
Why. Correction 6; addition J (Chua, IPS, 2024)
Before. Stewards "mostly to the performing arts and crafts"
After. Two food houses among the 2024 stewards; no hawker yet; NHB's Heritage Apprenticeship as the Fellowship's kin
Why. Addition E
Before. (UNESCO's marketing warning not cited)
After. §5.5 "What it must not become": decision 15.COM 8.B.6; consent, one dish, no stall signboards, a cap, a public register audited yearly, money to the family; §7.6: every campaign dollar reported against what reaches hawkers
Why. Addition D; the State is not said to have ignored the warning
Before. The at-risk count sent to four bodies, each to say how it will use it
After. One named recipient (the standing committee, or NHB until it exists); four new rows: a standing NEA–NHB–FMAS committee; joint owners for the report due 30 June 2027 with one shared Budget indicator; a hawker line in the next Heritage Plan; distributions, not medians; all zero or small cost
Why. Additions D, E, F and H
Before. Four-year ledger about 1.5 times the current rent support
After. … and less than half of the more than $50M the State gave more than 13,000 hawkers in 2020 (MSE, COS, 4 Mar 2021)
Why. Addition J
Before. (none)
After. "Brand stalls and the foreign-worker rule": a chain's stall may use its operator's quota, a family stall none (Chua, IPS, 2024; to be confirmed with MOM); a question, not a proposal
Why. Addition J
Before. Licence income $90,000 (Y3) → $3.0M (Y7); corpus real value 76 of 100; match shortfall $11.2M / $28.9M; corpus 105% / 129% / 86%; $126K carried, short from Year 6; $33.5M without HDB; at 20% no year short, 90%; at 30% $5.8M; at 40% $14.9M; Government licence match $2.0M
After. $60,000 → $2.0M, net of a third of licensees failing; 71 of 100 (20 short); $14.7M / $32.4M (§7.6.3 table from A4); 101% / 121% / 86%; $84K, short from Year 5; $37.0M; at 20% $0.06M short in one year, 87%; $6.9M; $16.0M; $1.4M
Why. numbers_A4.md §6: Chua's failure rate (hp_failure = 1/3, ASSUMPTION); levers and the Budget Book opener follow (one future in five)
Before. Proposal 1 precedent: the TWG; proposal 3: "To be priced"
After. Proposal 1 on ACPW Rec 9, kept opt-in on purpose (cash takings; the licence already enforces); proposal 3 priced on WICA's terms (Rec 6), with Malaysia's compulsory PERKESO cover since 1 Jan 2025 (38,358 hawkers contributing, Feb 2026); the ACPW's rejection of fixed-payout group cover (¶4.10.3) answered
Why. Addition C; H2 §3
Before. Co-contributions as a departure "from the principle of self-help"; three reasons; reason 3 "The match is the share that gatekeeper would pay"
After. Quoted as "departs from the principle of self-help", with the TWG's reason (¶54); reason 3 argued by analogy, conceding the Act's test; a fourth reason: the ACPW's co-contribution "part of the unique social compact in Singapore" (2022, accepted)
Why. Correction 5; addition C
Before. Livelihood dataset: earnings, hours, savings and injury
After. As distributions, with musculoskeletal pain (Nordic questionnaire), days closed for illness and the stated reason for every exit
Why. H2 note 4
Before. 19 Nov 2018 cited from the SG Parliament mirror (debate 9883); the CNA pieces undated; Dorsey (2025), Koh, T. (2022), Tan, K. P. (2021), NEA (2021; 2024) listed; "MAFF (2014), Ita-mae certification framework"
After. Official Report, Vol 94, No 85, pp. 13–35, Oral Answers; 2013 COS as Vol 90, Issue 13; CNA 19 Nov 2018 and CNA/TODAY 20 Oct 2018 dated; the four uncited entries removed; MAFF's 2016 cooking-skills guidelines (3,976 certified by Mar 2026); new reference groups for the record and for health
Why. Corrections 8 and 10
Before. Where the plan builds on others (§1.2, §3.2–3.6, §3.8 lines)
After. Adds the two panels, NHB's consent and IP practice, the monitoring promise, and the 2018–20 committee
Why. Additions A, D and F
Before. data/curated proposals 56; sources 244; glossary 119; tracker 20 indicators
After. Proposals 60 (four new §7.7 asks), stale licence and corpus figures and C wording fixed; sources 286 (46 added, 4 removed, MAFF and the 2018 sitting corrected); glossary 124 (five terms); tracker 21 (t05 partial, t21 Health and cover, t07 Flows, distributions in t08 and t19); day.json's FMAS and repairs notes
Why. The datasets follow the text (numbers_A4.md §7 for the tracker)
1.3 29 October 2026 164 changes · shared version
A grounding release. The library's claims were checked against their sources (314 checked; corrections made where the evidence differed, and what could not be confirmed is named in each volume), and the plan changes where the evidence says it should: the fee reset becomes a fee subsidy, the Hawker Apprentice Work Pass moves to Ideas for Later, apprentices are paid the Progressive Wage for the full 18 months, the Relief Corps is built on NEA's Joint Operator Scheme with a relief day the Employment Act allows, and Chinatown Complex starts after its pending repairs. Follow the Plate gains the middlemen (D.5.3), the plate over time and what a Regulars' List tier means in plates (D.4.2). On Uncle James's decisions of 29 October 2026: the Government gains one board seat with a casting vote, every use of it published within 30 days, and the Hawker Council's vetoes are internal to the Seed Bank, with a voting seat for FMAS; the Master Fellowship pays less each month for longer, a teaching year and a two-year stewardship ($90,750 a Fellow over three years); the Heritage Watchlist becomes the Regulars' List, one way on and no way down, in four tiers, with the at-risk count published only in aggregate; pilot centres are ranked by their likely count of heritage-worthy stalls; Hawkers Succession Scheme 2.0 becomes an extension of NEA's Scheme, and the endowment is the Hawker Endowment; the Family Name Royalty is capped at ten years. The four-year budget at three centres rises from $14,990,340 to $19,581,112.
Approved by Uncle James · tag v1.3 · release commit e9ba5e7
Before. about 6,000 of them in coffeeshops
After. about 5,800 of them in coffeeshops
Why. Aligned with C.7: 776 coffeeshops (HDB, July 2024) at about 7.5 cooked-food stalls each.
Before. Year 1 provisions roughly 2,400 subsidised relief-days
After. Year 1 provisions roughly 1,600 relief-days
Why. The Year 1 Corps line ($360,307) at the model's Year 1 day rate ($224) buys about 1,600 days, about 66 a member; the 1.2 figure did not match its own ledger.
Before. "close to half of tenderers would see their rent much higher than it is today"; "the end of the $1 rental bid stalls"; "only about 4% of stalls pay above the median assessed market rent of $1,250"
After. "close to about half of tenderers will see their rent much higher than it is today"; "the end of the $1 bid stalls that we have in the system today"; "only 4% of the 7,000 cooked-food stalls pay rent above the assessed market rent of $1,250"
Why. Quotations matched to the Hansard record of 13 November 2024 (Parliament 14, Vol 95, Sitting 147), read through Open Hansard; the 1.2 wording followed MSE's copy of the speech. Hansard also confirms Dr Koh's support for Edward Chia's amendment and the other quotations used.
Before. calling on the Government "to continue its support for hawkers by regularly reviewing its policies"; the amendment's text and any division unconfirmed
After. the resolution quoted from the Official Report, including "while enabling hawkers to earn a fair livelihood"; no Member recorded dissent
Why. The Official Report of 13 November 2024 (via SG Parliament) gives the motion as moved by Leong Mun Wai, Edward Chia's three amendments, the resolution as agreed, and the Speaker's call for dissent, which no Member recorded; Leong's closing speech says the PSP would support the amended motion.
Before. The 2024 proposals on the cheap plate, an academy, a foreign hand and the tender were credited to the PSP (and Leon Perera's 2021 motion) only
After. Louis Chua's speech of 13 November 2024 is credited too: Government-funded CHAS discounts, hawkers setting their own prices, a price-quality tender and a rent cap, a single independent academy, and a work pass for hawkers aged 60 and above
Why. Even-handed attribution: a cold political read found the WP's matching proposals uncredited (wp.sg, 'Growing Singapore's Hawker Culture', 13 Nov 2024).
Before. listed at Bronze when 20 regulars endorse it, or when the centre's hawkers' association nominates it
After. listed at Bronze when 20 regulars endorse it; the association may gather the names, but the names are the listing
Why. Consistency with the model, where Bronze always takes 20 named regulars (Uncle James, 29 Oct 2026).
Before. Riders and drivers get platform-paid CPF; 'most riders' have someone paying into their CPF
After. Every rider has work-injury cover; only riders born in or after 1995, and the 27% of older riders who opted in (about 26,500), have an operator CPF share; the 90,000 a month are mostly MediSave-only
Why. W2 #42, #43, #44 (MOM, 12 Jan and 5 Aug 2026)
Before. A MediSave-only hawker reaches 55 with no Retirement Account savings; CPF LIFE pays $0
After. Little or nothing: only MediSave above the Basic Healthcare Sum ($79,000 in 2026) flows into retirement savings, so for most such hawkers CPF LIFE pays little or nothing
Why. W2 #23 (MOM, 7 May 2022; MOM, 26 Jan 2026)
Before. MediSave paid in one annual lump, up to two years after earning; a quarter of the self-employed pay late and lose Workfare
After. Billed once a year after the tax assessment, payable by monthly GIRO; nothing collected as they earn; Workfare lost only if unpaid by CPF's deadline
Why. W2 #18, #21 (CPF, 5 Jun and 22 Oct 2026)
Before. Nothing enforces MediSave; a quarter of stallholders in arrears
After. The SFA stall licence is not renewed with MediSave arrears unless the hawker is on GIRO; the quarter is for all self-employed persons, and hawkers' own rate is probably lower and unpublished
Why. W2 #19, #20 (SFA; CPF, 4 Nov 2025)
Before. No income-loss support when illness shuts the stall
After. No injury cover, sick pay or income-loss insurance of their own; means-tested ComCare Short-to-Medium-Term Assistance and private Prolonged Medical Leave policies (since 2018) exist
Why. W2 #33 (MSF; NTUC, 7 Sep 2018)
Before. Proposal 1, MediSave as you earn, closes 'the annual lump-sum bill, and the arrears that cut off Workfare'; 'the annual bill and its arrears disappear'
After. Re-framed around smoothing cash flow: the bill is paid in small sums as the takings come in; the licence already enforces payment, so fewer arrears are no longer claimed
Why. W2 new evidence on proposal 1
Before. MRSS precedent stated without scale; the 2018 Workgroup cited only for daily cash
After. MRSS: $456M to more than 250,000 members for 2025, and persons with disabilities under 55 from 2026; the Workgroup's objection that Government co-contributions 'depart from the principle of self-help' is stated and answered in C.8.1
Why. W2 #26, #38
Before. Pooled cover presented as if nothing exists; the Workgroup's daily-cash recommendation as precedent
After. Pooled cover is additive to ComCare and to the medical-leave policies insurers already sell; the pilot asks why so few hawkers buy them; the $750 premium marked an assumption
Why. W2 new evidence on proposal 3; W4 #59, W5 #46
Before. A relief cook covering a hospitalised hawker breaches both rules; the nominee route alone
After. Names NEA's Joint Operator Scheme (since 2012) and the nominee route for illness (Dr Koh, 10 Mar 2025); the Corps is built on them and asks for speed and a short-cover variant
Why. W5 #37, #38, #36
Before. Sources notes name where the sources are listed
After. Each sources note adds 'What we could not confirm': the ledgers' unverifiable and unreachable claims for that volume, named plainly
Why. Decision 7: every unverifiable claim softened or named in that volume's sources note
Before. The assistant he employs, most riders who deliver his food and the office worker ... each have someone else paying into their CPF
After. The assistant and the office worker do, and so do the younger riders and the one in four older riders who opted in; every rider has injury cover from his platform
Why. W2 #43
Before. Employer, 17%
After. Employer, 17% (less above 55); rider's share 'if born 1995 or later or opted in'
Why. W2 #39, #43
Before. No one pays a retiring hawker for his goodwill ... keeps his recipes well enough to cook from, gives a new cook an income floor ..., gives any hawker a day off
After. No scheme pays ... – the few private sales are one-off deals – ... keeps his recipes in a standing archive good enough to cook from, tops up a new cook's own income for two years, sends a relief cook so a hawker can take a day off without closing
Why. W3 #42, #44, #49, #32
Before. Two rows are empty; repairs and exhausts belong to the landlords, 'Left to landlords'
After. One row is empty (knowing who is at risk). Repairs and exhausts: NEA and the Town Councils repair and upgrade (R&R every 6–8 years, HUP 2.0 up to $1bn); the stallholder keeps his own duct; the gap is a stall's own equipment while the HPG is between phases. Government 'partly'; the plan's mark 'Not ours to fund', no section, no payer
Why. W1 #40; W2 #7
Before. A ballot for stalls was proposed in Parliament in 2024 and turned down, and §2.2 answers the Government's objection
After. Proposed by the PSP's two NCMPs, Leong Mun Wai and Hazel Poa; the Government argued against it, and §2.2 answers each of its four objections
Why. W3 #2, #3, #7; W5 #6
Before. 2.8% HDP graduates still at a stall
After. 2.8% HDP trainees running a stall: 16 of 566; 29 started one
Why. W1 #26
Before. A Relief-Hawker Corps so a cook can take a day off; 'days off'
After. Built on NEA's joint-operator and nominee routes, so a cook can take a day off without closing
Why. W3 #49; W5 #37, #38
Before. A table-cleaning fee re-priced for tray return; fee income NEA or a Town Council gives up
After. Cleaning contracts re-tendered for tray return at the pilot centres, with a $200-a-month table-cleaning fee subsidy as a named Budget cost; nationally up to about $16.6M a year, all the Government's
Why. Decision 1
Before. A 50-pass Apprentice Work Pass pilot among the decisions asked of Government
After. No longer asked for: asked on 10 Sep 2026 why work permit holders may work in coffeeshops but not in hawker centres, the Government left its rules as they are, citing heritage, and the idea waits among the ideas for later (§7.9)
Why. Decision 2
Before. No ask on designated suppliers
After. No compulsory designated suppliers at operator-run centres, among the decisions asked of Government
Why. Decision 6
Before. A paid apprenticeship
After. An apprenticeship paid at the Progressive Wage for a stall assistant; ES.7 says plainly that the bill is higher than in the last version, because of the apprentice pay and the Fellows' CPF
Why. Decision 3; W4 #47
Before. All three start at two Lighthouse centres ...
After. Year 0 confirms the dates of works at both with NEA and the Town Council; Chinatown Complex starts only after its delayed repairs and redecoration
Why. Decision 4
Before. In one minute boxes and the §7.7 short form: '$1,500 a month' apprentices; relief cooks 'so a hawker can take a day off'; the fee re-priced and its income given up; a 50-pass Work Pass
After. Apprentices at the Progressive Wage ($2,220 from July 2026, rising to $2,500 by July 2028); relief cooks built on NEA's routes, a day off without closing; the fee subsidy, cleaning contracts re-tendered, no compulsory designated suppliers; the Work Pass among the ideas for later; Chinatown's start after its repairs; MediSave as you earn about cash flow; the margin-spread line in 'Going deeper'
Why. Decisions 1–4 and 6; W2, W3 #49 (minute_requests_P.md)
Before. Part II contents: '… Relief Corps · Work Pass · Security Package …'
After. '… Relief Corps · Work Pass, deferred · Security Package …'
Why. Decision 2
Before. Levers: three NEA rents 'about $1,200 assessed, $1,450 paid'; 'the fee's cost' as income the fee collector gives up, the operator bearing its share; delivery 15% instead of 22%; the Corps and named deputy without the Joint Operator Scheme; the matching pool without CMF terms
After. $1,250 the system median and the median AMR, $1,450 the comparable-centre median, $1,800 the winning bid; the fee subsidy, a Government cost against a contractor's price, the 6 Jul 2021 reply cited, a pass-through at operator-run centres; 15% instead of 30%; the circles' saving shared between supplier and cook; the tied-supplier ask; the Corps on the JOS, the deputy a short-cover joint operator; CMF caps and no double matching; MediSave as you earn a cash-flow lever
Why. Decisions 1 and 6; W1 #5; W3 #24; W4 #30; W5 #9, #37, #38, #61; W6 #28
Before. Openers: the ask without the fee subsidy; the ledger total without comment; 'no one pays ... gives any hawker a day off'
After. Budget Book: the fee is now a subsidy and the Work Pass has left the ask; the bill is higher because of apprentice pay and the Fellows' CPF; Chinatown's start after its repairs. Who Already Helps: 'no scheme', the private sales, the standing archive, a day off without closing, Kopitiam's HawkerBoss
Why. Decisions 1–4; W3 #32, #42, #44, #49
Before. $14.99M ($14,990,340) over Years 0–3, about 65% to hawkers; add-ons $3,898,880 and $3,068,935; tax forgone about $5.3M; downside Year 3 about $4.3M
After. $18.86M ($18,863,142), 72% (72.1%) to hawkers, $3,872,802 more, said plainly; add-ons $5,304,561 and $4,191,829; tax forgone about $5.7M (donors at 23%); downside Year 3 about $5.4M
Why. Decision 3 (apprentice pay); W4 #47, #55 (numbers_M.md)
Before. Levers and openers at the 1.2 model: break point $720; Corps leverage 2.65; $9.42 an hour; $8.0M stretch; $3.50M a year from Year 4; circles $435 a month
After. At the 1.3 model: $610; 2.17; $9.69 (with the package $15.60; 66% → 26% below the benchmark); $7.9M; $4.05M; $407; bench sizes from the model (15→23, 4→6, 8→12)
Why. Model inputs corrected (rent $1,250, delivery 30%, wage shock with CPF, Corps at the PWM cook rate): W1 #5, W4 #11, #30, #31, #49; numbers_M.md
Before. Fee reset: NEA re-prices the table-cleaning fee, $200 a month less; fee income forgone, an operator bearing its own share
After. A $200-a-month table-cleaning fee subsidy paid by the Government at the pilot centres, after NEA's temporary support (Sep 2026–Feb 2027), tested by re-tendering the cleaning contracts; a pass-through at operator-run centres
Why. Decision 1; the Minister said on 6 Jul 2021 that no fall in cleaning costs was expected, and NEA bars operators from profiting from pass-through charges (W5 #9, W1 #30, W1 #31)
Before. Ballot credited to Leong Mun Wai alone, 'at a flat rent'; one Government objection answered
After. Leong (turnover rent with a floor) and Hazel Poa (ballot, pre-determined rent) both credited; Dr Koh's four objections each answered, by the qualification gate, the bands, no bidding inside a band, and a concession on the 4%
Why. W3 #2, W3 #3, W5 #6
Before. No domestic precedent cited for fixed rents without bidding
After. Operator-run centres cited: rents fixed in the operator's contract, capped at valued rents since 2019, not subject to bidding; their 8% early terminations against 3% given with MSE's caveat; Bukit Ho Swee's 1966 ballot
Why. W1 new evidence; W1 #29, W1 #42, W5 #7
Before. Median AMR ≈$1,200; $1,450 the median rent non-subsidised stallholders paid
After. Median AMR $1,250 (13 Nov 2024); system median rent about $1,250, stable for a decade; $1,450 the 2025 median at centres comparable with the operator-run ones
Why. W1 #5, W3 #24, W5 #11, W5 #12
Before. Chinatown Complex's band tier unstated; GST on rents not mentioned
After. Each pilot centre's tier set from its own tender record (Chinatown Complex's median bid $1,408, Aug 2024–Jan 2025); rents and conservancy quoted before GST
Why. W1 new evidence; W6 #14
Before. §2.8 Hawker Apprentice Work Pass: a MOM pilot of 50 passes for one foreign apprentice a stall
After. §2.8 a short stub: deferred, with reasons; the full idea in §7.9, crediting Leong Mun Wai (13 Nov 2024), Dr Koh's 'not inconceivable', the Government's refusal of 10 Sep 2026 and the LTVP route (about 170, then about 300)
Why. Decision 2 (W3 #12, #13, #14; W5 #42, #43)
Before. Apprentice Work Pass pilot row in the ask
After. Row removed; 'No new work pass' added to what we do not ask
Why. Decision 2
Before. Relief notice and named deputy as new NEA rules; 'a relief cook ... breaches both rules'
After. Built on the Joint Operator Scheme (since 2012) and the nominee route: a relief notice making a Corps cook a joint operator for the absence, without the four-hour duty; the named deputy as a short-cover joint operator
Why. W5 #37, W5 #38
Before. Corps member earns $190/day; benches start at 15 cooks and 4
After. Day rate indexed to the PWM cook rate, employer CPF included; benches 15→23 and 4→6, Tekka 8→12, as in the ledgers
Why. W4 #49; W5 bench sizes disagree across §2.7, D.11 and B.8
Before. The platform takes 'about a fifth of the sale'; group rate 15% 'against the model's 22%'
After. About 30% plus GST, GrabFood's published fee; most of it pays riders; Deliveroo left on 4 Mar 2026; 15% a target to negotiate against the published 30%
Why. W4 #30, W4 #31, W6 #11
Before. 'Suppliers' margins fall by the same 3% ... The money moves from the middle of the chain to the cook'
After. A lower cost of serving many small buyers, shared between supplier and cook; no profiteering claimed; the genuine tolls named (platforms, tied suppliers, the 2018 chicken cartel); a competition-law guardrail; an ask that NEA bar compulsory designated suppliers
Why. Decision 6; W6 #22, W6 #28; W5 #49
Before. No ask on tied suppliers
After. New row: no compulsory designated suppliers at operator-run centres (NEA guidelines, zero cost)
Why. Decision 6
Before. No indicator of supply-chain margins
After. A margin-spread indicator on the Plate Ledger: retail and wholesale staple prices against the food import price index, quarterly
Why. Decision 6
Before. Apprentices receive $1,500/month throughout
After. Paid the PWM for a food/drink stall assistant: $2,220 (Jul 2026), $2,360 (Jul 2027), $2,500 (Jul 2028), plus employer CPF, for all 18 months; the budget rises
Why. Decision 3 (W4 #48, W5 #30)
Before. Graduate retention ≥60% at 5 yrs (HDP baseline: 2.8%)
After. ≥70% at 5 yrs, set against two HDP baselines: conversion 29 of 566 (5%), retention 16 of 29 (55%)
Why. W1 #26. TARGET CHANGE FOR UNCLE JAMES TO CONFIRM: ≥60% raised to ≥70%
Before. 2.8%: HDP graduates still operating a stall (16 of 566); 'a handful operating'
After. Of 566 who finished classroom training, 16 run a stall; 29 started one, 16 still operate
Why. W1 #26
Before. Four interventions
After. Adds the WDA–NEA Master Hawker Programme (2013–14, 13 trainees, operations only, not recipes), as reported by Makansutra
Why. W3 outside-domain note
Before. 'No repair fund and no ducting programme anywhere'; repairs and exhausts 'the landlord's'; HPG 'closed'
After. The State repairs and upgrades centres through R&R every 6–8 years and HUP 2.0 (Mar 2025, up to $1bn, ventilation included); the stall's duct is the stallholder's legal duty; the gap is stall-level equipment and duct upkeep while the HPG is between phases; new §1.3 row
Why. W1 #40, W2 #7
Before. CDC tranches $300 and $500; NEA does not require budget meals; personal operation with 'flexibility'
After. Adds the $300 tranche for Jan 2027; the SEHC value-meal tender condition; the Joint Operator Scheme and nominee route; PHC dishwashing co-funding
Why. W2 #13, W5 #24, W5 #37, W2 #8
Before. Goodwill on the HSS rail 'within existing scheme parameters'; 'coffee money' changes hands
After. NEA must amend the HSS terms to allow a disclosed, capped goodwill payment for recipes, training and name, never the location; Kay Lee ($4M, 2014) and Sin Kee as precedents, with the food-drift lesson
Why. W5 #18, W5 #56
Before. Extend eligibility from 15-year veterans
After. Beyond the veterans of about 15 years the pilot was designed for (NEA's page no longer states a minimum)
Why. W2 #5, W5 #16
Before. Royalty to 'the master's family', unconditioned
After. No royalty until the name is a registered trade mark with a named holder and the heirs' shares are set; why a name royalty is not a sublet
Why. W5 #57
Before. Band rent minus a 30% heritage subsidy; Heritage Excellence 'rent rebate'
After. A 30% heritage rebate paid to the graduate in cash from the package, never off the rent; the Excellence rebate a cash rebate sized to rent
Why. W5 #34: conflict with §2.2's 'no rent subsidy from the programme'
Before. No mention of Kopitiam's HawkerBoss
After. HawkerBoss named as the nearest analogue; Kopitiam as a partner; the package tops up only what HawkerBoss does not pay
Why. W3 #32
Before. Makan tranche on CDC rails as they stand; NEA requires no budget meals
After. Two changes asked of MCCY and the PA (a hawker-and-coffeeshop-stall tag, an eligibility list); the SEHC value-meal tender condition made opt-in and compensated at the pilot centres; PSP's 2024 call credited; Bukit Canberra: the operator removed the clause
Why. W5 #24, #25, #28; W3 #18, #22
Before. Roughly 400 HDB-rented and 400 private coffeeshops; the landlord's side unpriced
After. 374 rented from HDB and 402 private (776, Jul 2024); the rent a private landlord gives up (about $2,250 a month) stated, private uptake a hypothesis
Why. W1 #36, W5 #23
Before. About 400 + 400 coffeeshops; median asking rent near $3,400 (source unnamed in the text)
After. 374 + 402; the $3,400 named as a listings median (Stall.sg via DollarsAndSense), beside commercial guides' $4,000–10,000
Why. W1 #35, W1 #36
Before. Chinatown Complex's major R&R 'planned for 2026, so ... finished before Year 1'
After. Reported for March 2026, not found to have happened, last R&R ended May 2019; Year 0 confirms dates with NEA, HDB and the Town Council and Chinatown's start is timed after its works; Geylang Serai refurbished in 2024
Why. Decision 4 (W1 #15, #16, #46)
Before. Town Council collects both fees; an operator's fee income given up is its own; Tekka 'North Indian, Pakistani and Chinese'
After. Town Council collects the conservancy charge; operator-run rents fixed in NEA's contract and the fee a pass-through; FairPrice centres' tenancies on NEA's behalf; Tekka's Indian-Muslim cluster named
Why. W1 #3, #4, #17, #29, #30
Before. $10M matching pool 'sought first from MCCY's Cultural Matching Fund'
After. The CMF one source of it, on its terms: cash only, yearly windows, $300,000 a window for general use, no double matching with other government grants; sector class and corpus use to be confirmed with MCCY
Why. W5 #61
Before. 'as a platform now shares the CPF cost for a rider'; 'The rider got CPF'
After. For riders born in or after 1995 and the roughly 27% of older riders who opted in; every rider has work-injury cover
Why. W2 #43
Before. ≈$750 a head for group CI, unqualified
After. Marked ASSUMPTION until an insurer quotes, with the sum assured, ages and pre-existing-condition terms to be stated
Why. W4 #59, W5 #46
Before. Council vetoes stated; conflicts rule bars only Fellowships
After. Vetoes written into the constitution as reserved matters needing the Council-class members' consent; hawker directors' other benefits only on the same terms as every eligible stallholder
Why. W5 #63
Before. '8 of the 55 ... were flagged'; maintenance rate of 10% unexplained against 15–25% at risk; UNESCO report 'due 30 June 2027'
After. The 8 of 55 marked as the authors' own working; why 10% a year differs from the 15–25% stock; the UNESCO date attributed to UNESCO's state page
Why. W1 #24, W4 #58, W5 #52
Before. Documentation precedents without Temasek Polytechnic's book
After. Credits Singapore Hawker Classics Unveiled (2015)
Why. W3 #44
Before. Exhausts, ducting and stall repairs 'the landlord's and the hawker's own business'; 70 in 120 aged 55+ unqualified
After. Centres' repairs through R&R and HUP 2.0; the fee subsidy is outside the ledgers; 70 in 120 marked ASSUMPTION
Why. W1 #40; W1 unverifiable table
Before. Government's exposure includes 'the table-cleaning fee income NEA gives up'
After. The fee subsidy listed among the Budget costs
Why. Decision 1
Before. References list as of 27 Oct 2026
After. Ten source groups added (rents and fees; repairs; the 2021 and 2024 debates; relief and manpower; training and the HDP; more help; vouchers; Buying Circles and the chain; goodwill and the endowment; riders), with the ledgers' URLs; context paragraph corrected
Why. Brief: cite each new fact and add each new source
Before. Four-year programme at three centres $14,990,340 ($9,727,500 reaching hawkers, 64.9%); Years 1–3 $3,588,685, $5,013,200, $5,553,500
After. $18,863,142 ($13,600,302, 72.1%); Years 1–3 $4,412,571, $6,448,967, $7,166,649; the rise of $3,872,802 (26%) explained: apprentices at the PWM wage with CPF (~$3.31M), Fellows' CPF (~$0.25M), the indexed and lawful Corps day (~$0.31M)
Why. Decision 3; W4 #47, #48, #49 (numbers_M.md)
Before. Add-ons: Old Airport Road $3,898,880, Maxwell $3,068,935; four centres $18,889,220, five $21,958,155; a further centre $2,583,125; the Lighthouses alone $12,500,680; per-centre A+B $6,082,685 / $2,540,935 / $2,489,660
After. $5,304,561 and $4,191,829; $24,167,703 and $28,359,532; $3,460,004; $15,568,579; $8,453,167 / $3,238,352 / $3,294,563, with shares and Year-4 costs re-run
Why. numbers_M.md
Before. Apprentice stipends 20 × $1,500 × 12, 32 × $18,000, 36 × $18,000; Master Fellowships n × $72,000; Corps at $12,500 a member
After. Apprentice wages 20 × $50,684 and 32/36 × $52,650 (18 months at the PWM assistant wage with employer CPF, budgeted in the starting year); Fellowships n × $77,250, CPF included; Corps at the PWM cook rate
Why. Decision 3; W4 #47, #49
Before. Flat $72,000 a Fellow, 'plus CPF contributions on stipend'
After. Flat $77,250, with up to $5,250 of employer CPF on the stipend costed
Why. W4 #47
Before. Break points $720, $2,590 (package), $2,300 (no deputy); 62–75% / 22–34% across $1,200–1,800; loss above $4,500; successor keeps $1,580; protection binds from $1,920
After. $610, $2,450, $2,150; 66–77% / 26–37% across $1,250–1,800; loss above $4,400; $1,510; $1,810
Why. Rent $1,450 → $1,250 and delivery 22% → 30% in the model (numbers_M.md)
Before. Delivery group rate 15% counted; circles add $435 a month, $5,124 a year; break points $1,660 → $1,900 / $1,760 / $1,760; delivery $310 a month
After. 25% group rate counted against the 30% list rate, 15% kept as the target and shown as a sensitivity (about $1,760); circles add $407 a month, $4,791 a year; $1,540 → $1,780 / $1,640 / $1,610; delivery $424 a month
Why. W4 #31; coordinator's instruction; numbers_M.md
Before. Middlemen evidence pointed at Follow the Plate generally
After. Points to the new section, D.5.3
Why. Decision 6
Before. Seventeen indicators
After. Eighteen indicators (the margin spread added)
Why. Decision 6
Before. Day rate indexed (no figure); named deputy moves the break point $1,660 → $1,950
After. About $212 a day in 2026, $224 from July 2027, $236 from July 2028, for a relief day of at most 12 hours with overtime beyond 8; $1,540 → $1,830
Why. W4 #49; numbers_M.md
Before. Bridge gaps $2,568,685 / $3,163,200 / $3,153,500, about $8.9M (Lighthouses $6.4M); tax forgone ≈$5.3M; $1.26 public per private dollar; endowment keeps real value in 89 of 100 futures; gifts meet $1.5M a year; Y10 draw ~$1.9M
After. $3,392,571 / $4,598,967 / $4,766,649, about $12.8M (Lighthouses $9.5M); ≈$5.7M (donors at 23%; $5.6–5.9M at 22–24%); $1.11; 78 of 100 (13 with a short year); $2.1M; ~$1.8M
Why. W4 #55; numbers_M.md
Before. Stretch match ≈$8.0M; base shortfall Years 4–10 $8.8M, stretch $26.8M ($31.7M without HDB); base short from Year 7; ~$1M a year of gifts set aside; corpus Y20 118% / 160% / 86%; HDB lever ≈$1.8M, under-55 part ≈$0.9M
After. ≈$7.9M; $10.9M, $28.6M ($33.2M); from Year 6; ~$0.7M; 106% / 133% / 86%; ≈$1.7M, ≈$0.84M; the year-by-year table re-run
Why. numbers_M.md
Before. Lighthouses' Year 3 $4.3M, from Year 4 $2.88M, four years $12,500,680; three centres from Year 4 $3.50M; maintenance dial: 20% above target, 30% just keeps value, 40% $0.8M short
After. $5.4M, $3.31M, $15,568,579; $4.05M; 20%: $1.6M short over Years 4–20, corpus 93%; 30%: $4.3M short over Years 4–10; 40%: $12.8M
Why. numbers_M.md
Before. Expected first loss ≈$3,400 (worked example), ≈$6,500 averaged
After. ≈$3,200 and ≈$5,900
Why. numbers_M.md (succession_lab)
Before. HDB forgone tender rent up to about $1.8M a year
After. Up to about $1.7M a year (374 HDB-rented coffeeshops)
Why. W1 #36; numbers_M.md
Before. On One Page strip: $14.99M programme
After. $18.86M
Why. numbers_M.md
Before. Tendered NEA stall at $1,450, 'what sitting tenants actually pay'; quoted rents $1,200 AMR, $1,450, $1,800
After. Tendered stall at the $1,250 median across NEA's centres (also the median AMR); $1,450 kept as the 2025 median at centres comparable with operator-run ones; quoted rents $1,250, $1,450, $1,800
Why. W4 #11, W1 #5, W1 #32, W6 #26: $1,450 is a comparable-centre median (MSE, 8 Sep 2026); the system median is about $1,250
Before. Delivery commission 22% (Monte Carlo 15–29%), from a vendor blog
After. 30%, GrabFood's published fee before GST, varied 25–35%; Deliveroo's exit noted
Why. W4 #30 (high): Grab publishes 30% plus GST
Before. Buying Circles' delivery at 15% instead of 22%
After. Counted at a 25% group rate against the 30% list rate; 15% kept as the circles' target, shown as a sensitivity
Why. W4 #31: 15% is unsupported and half the published rate; test 20–25% as the central case
Before. Table-cleaning fee reset booked as fee income the collector gives up; operators bear 45% at SEHCs
After. A $200-a-month fee subsidy the Government pays to the cleaning contractor at pilot-centre stalls, all public money, an upper bound tested by re-tendering; landlord take unchanged
Why. Decision 1; W1 #30, W1 #31, W5 #9; MSE 6 Jul 2021
Before. 805 HDB coffeeshops (405 rented, 400 sold); 3,000 HDB-rented coffeeshop stalls; 14,400 stalls in all; HDB lever up to $1.8M
After. 776 (374 rented, 402 private, July 2024); 2,800 stalls; 14,200 in all; HDB lever up to $1.68M; tender receipts about $215M, an upper estimate
Why. W1 #36, W6 #24, W4 #20
Before. Relief cook paid a fixed $190 a day; Corps member $12,500 a year
After. PWM Cook (QS) rate for a relief day of at most 12 hours, the 4 beyond the normal 8 at 1.5 times, with 17% employer CPF: $212 (2026), $224 (Year 1), $236 (Years 2–3); member $14,753 then $15,539
Why. W4 #49: index to the PWM cook rate (MOM, 16 Mar 2026)
Before. Apprentices costed once at $18,000 ($1,500 a month, 12 months)
After. PWM stall-assistant wage for the full 18 months plus 17% employer CPF, budgeted when the cohort starts: $50,684 (Year 1 start), $52,650 (Years 2–3)
Why. Decision 3; W4 #48 (high)
Before. Master Fellowship $72,000 'plus CPF on the stipend', CPF not costed
After. $77,250 a Fellow: employer CPF of $5,250 on the $42,000 stipend at the 12.5% (above 60 to 65) rate, a ceiling
Why. W4 #47; CPF Board 2026 rates
Before. Years 0–3 at three centres $14,990,340; two Lighthouses $12,500,680; from Year 4 about $3.50M a year
After. $18,863,142; two Lighthouses $15,568,579; from Year 4 about $4.05M a year; corpus at Year 20 133% of target (was 160%)
Why. Decision 3 and the costing corrections above
Before. Top-bracket donors' tax rate 20%; tax forgone about $5.3M over the campaign
After. 23%, the middle of the 22–24% top bands; about $5.7M ($5.6–5.9M)
Why. W4 #55: the plate model's own income-tax bands
Before. The circles move about $12.0M a year 'from suppliers' margins' to hawkers
After. Circle members pay suppliers about $11.8M a year less: a lower cost of serving many small buyers, shared between supplier and cook
Why. Decision 6; W6 #28; margins tracked import prices 2019–25
Before. Seventeen indicators in the Plate Ledger
After. Eighteen: adds the margin spread (staple prices at wholesale and retail against the food import price index, quarterly)
Why. Decision 6
Before. Eighteen tracker indicators (3 published, 9 in part, 6 not at all)
After. Nineteen: adds floor costs by type of centre (rent, conservancy, table cleaning and dishwashing, before GST), in part (MSE's 2025 medians)
Why. D.5.4: the $2,750 against $2,170 floor-cost gap at operator-run centres (MSE, 8 Sep 2026) should be tracked, not argued about
Before. NEA re-tenders the pilot centres' cleaning contracts; who holds the contract at a Town Council centre to be confirmed; the Town Council collects the table-cleaning fee
After. The contract's holder re-tenders, convened by NEA: at HDB hawker centres (Chinatown Complex and Tekka among them) NEA's notices say the table-cleaning service 'is managed by the respective Hawkers Association'; at Geylang Serai (NEA) Year 0 confirms the holder
Why. NEA tender notices, Dec 2022 and Aug 2024, Note 1(d), confirmed on 29 Oct 2026; the note covers HDB hawker centres, not NEA's own; Aug 2026 notice names only 'the table-cleaning contractor'. The subsidy, its amount and its payer are unchanged
Before. The fee follows the contractor's price; the cleaners' PWM rose from $1,312 to $2,080
After. Adds: at four NEA-managed centres the fee was the same before GST in Dec 2022 and Aug 2024 ($402 at Amoy Street, $360 at Chomp Chomp) while the cleaner's PWM rose by a third ($1,312 to $1,740); a re-tender may find no saving
Why. Derived from NEA's tender notices (fees printed with 7% and 9% GST) and MOM
Before. There is no Singapore estimate of how diners respond to hawker prices
After. One early, indicative estimate: Poh (2026), a stated-preference survey of 90 diners, not random; elasticities 2.81 (18–24), 1.01 (45–60), 3.92 (60+, nine people), 2.09 on low meal budgets against 1.1–1.2; the model's −0.5 is unchanged
Why. Poh, Decoding the Consumers of Hawker Food (Scholarly Review Journal, Fall 2026); no model parameter changes
Before. MTI 2015 cited for its pass-through coefficients only
After. Adds its stall-level panel (2012–14): about two in three stalls that raised prices had no rent change; subsidised and non-subsidised stalls priced alike; higher pass-through in the CBD; one more similar stall, prices about 8.4% lower; publishing prices was among its measures
Why. MTI, Economic Survey 1Q2015, checked against its text; cited at its current URL
Before. The relief cook paid the PWM cook rate for the stall's whole 13-hour day, at straight time, with employer CPF (about $197 in 2026)
After. A relief day of at most 12 hours: 8 normal hours at the PWM Cook (QS) rate and 4 at 1.5 times, with 17% employer CPF, about $212 in 2026, $224 from July 2027 and $236 from July 2028; on a 13-hour stall day the stall's own assistant does the first hour's prep, and a stall with no assistant opens an hour later
Why. The corps member is an employee: the Employment Act (Part 4) caps the day at 12 hours, sets normal hours at 8 a day for a week of more than five days and requires overtime at 1.5 times (MOM). Review B1
Before. so the day rate never falls behind the wage it competes with
After. Removed; the hawker co-pays $80 a day and the programme pays the rest of a lawful day
Why. The claim did not hold at straight time over 13 hours. Review B1
Before. Corps member $13,699 (Year 1) and $14,429 (Years 2–3); Corps lines $340,281, $653,873, $711,589; the Corps' share of the rise about $0.19M
After. Corps member $14,753 and $15,539; Corps lines $360,307, $694,943, $757,099; the Corps' share about $0.31M of a $3,872,802 rise; the Relief Corps' leverage 2.17 (was 2.45); national package $78.7M a year, leverage 1.06
Why. The model re-run with the lawful relief day (plate_model.py, funding_model.py; both parity tests pass)
Before. Corpus at Year 20 135% of target; real value kept in 79 of 100 futures, a short year in 11; gaps about $12.6M (Lighthouses $9.4M); from Year 4 about $4.00M a year
After. 133%; 78 of 100, a short year in 13; gaps $3,392,571 / $4,598,967 / $4,766,649, about $12.8M (Lighthouses $9.5M); from Year 4 about $4.05M a year
Why. Follow the Gift re-run on the new ledgers
Before. From Year 4 … no year runs short
After. At the central assumptions no year runs short; in about one future in eight some year does
Why. In the Monte Carlo 12.7% of futures have a short year. Review S12
Before. Employer CPF on the stipend, at most 12.5% … up to $5,250; FLAT $77,250 FOR EVERY FELLOW; a ceiling
After. Employer CPF at 12.5%, the 2026 rate for ages above 60 to 65: Fellows are veteran hawkers over 60, and one under 60 would cost more (16–17%, $6,720–7,140); $77,250 a Fellow over 60
Why. 12.5% is not a ceiling below 60 (CPF Board, rates from 1 Jan 2026). Review S1
Before. Grace Fu kept work permit holders out of hawker centres "to preserve the local identity and heritage of our hawker culture"
After. Asked why work permit holders may work in coffeeshops but not in hawker centres, the Minister left the rules as they are: "Hawker centres are a unique part of Singapore's cultural heritage and it is important to preserve the local identity" (MSE, 10 Sep 2026); MOM's rule "To preserve the local identity of our hawker culture and heritage" (25 Feb 2026); both sources added
Why. Each quote now matches its source, and each claim says only what the reply says. Review S2
Before. Retirement savings: Fills, under 55 (§2.11), paid by the State and gifts
After. Costed, for later (§2.11), no payer until a Budget decides
Why. The match is costed and left for later, not asked for. Review S3
Before. 16 of 566 HDP trainees run a stall; 29 started one
After. 16 run a stall in a hawker centre; 29 started one there (stalls elsewhere are not counted)
Why. The reply counts hawker-centre stalls only (MSE, 2 Jul 2024). Review S13
Before. Grounded: every claim checked against its sources and corrected where wrong
After. Grounded: the library's claims checked against sources (314 checked; corrections made where the evidence differed)
Why. Each volume names claims that could not be confirmed. Review S14
Before. no one keeps a list; which no one counts today; Fills a gap no one fills; No one keeps a hawker's recipes; No one finances or guarantees them; No one publishes those
After. Government keeps no list; which Government does not count and we found no one else who does; Fills a gap (none found); No scheme keeps …; No scheme finances …; No one publishes them regularly; only one-off medians and cost shares exist (MTI 2015; NEA 2022; MSE 8 Sep 2026)
Why. Say what the evidence supports. Review S15
Before. operators who mandated loss-making meals, harvested sales data, and used it against their own tenants
After. operators who applied loss-making value meals and set data-sharing and app-discount terms that many stallholders resented, as reported (the value meal was NEA's tender condition: MSE, 24 Sep 2025)
Why. No source for data used against tenants. Review S16
Before. Deliveroo left Singapore on 4 March 2026
After. Deliveroo went offline in Singapore after 4 March 2026
Why. It stayed live until 4 March. Review nit 8
Before. HawkerBoss … in Kopitiam and Foodfare outlets; Kopitiam and Foodfare as partners
After. in Kopitiam outlets; Kopitiam as a partner
Why. The 10 May 2022 release names Kopitiam outlets only. Review nit 6
Before. apprentice stipends; $31.7M of public money (anchor, match, tax forgone and operating match); about $18.7 million
After. apprentice wages; NHB's half of the Year 0 research grants ($321K) listed; about $18.86 million
Why. Review nits 9, 11, 15
Before. the one in four older riders; between HPG phases, nothing funds it; I agree with the PSP; one joint operator; at any sane price; nearly half of Singapore's hawkers; Chinatown's start timed after its works
After. more than one in four; the HPG co-funds new equipment, never repairs; I agree that stalls should be allocated by ballot (and with the proposal); a joint operator; clause removed; about two in five; the ledgers assume the works are done before Year 1 begins in July 2027
Why. Review nits 13 (Figure D.6), 14, 16, 17, 18, 19, 21 and the unsourced §2.9 clause
Before. No source for the Employment Act's hours or the CPF rate table
After. MOM, hours of work and overtime; MOM, part-time overtime; CPF Board, contribution rates from 1 Jan 2026; MOM, written answer of 25 Feb 2026
Why. Each new fact cites its source
Before. All of it is meant for every stallholder, celebrated or not
After. Part II is the floor for every stallholder, carried mostly by the State as rules and a fee subsidy; Parts III–IV are the craft tier, paid for by gifts and open to any stall that passes the test
Why. The two tiers are paid for differently; a political and incidence review (29 October 2026) found the opening blurred them.
Before. Heritage-worthy is a proxy (MICHELIN, awards, heritage records, padded)
After. The proxy sizes the budget only; from Year 1 Fellowships follow the Watchlist's open at-risk tests (§3.6), never fame alone, and the Seed Bank publishes each year who received them by centre, cuisine and community
Why. The proxy counts the noticed and tilts: Tekka's heritage-worthy count is 7.6% of its stalls against 11.9% (Chinatown Complex) and 12.7% (Geylang Serai).
Before. Makan Vouchers: cost-of-living relief and heritage stewardship in one visible move
After. Cost-of-living relief for the households that need it, with a small broad lift for every hawker (about $360 a stall a year in the model); not heritage stewardship
Why. By the model's own numbers (results.json: $4.58M of hawker household income on $66M, 12,700 stalls) the tranche is mainly relief for diners. Same fix in data/curated/proposals.json.
Before. (no statement of who the money reaches)
After. Who the money reaches: of $18,863,142, about half to the 44 heritage-worthy stalls, a seventh to all 408 pilot stalls, a third overhead; 82% gift money; the State's lines with the tranche about $234M, 85% vouchers; the ledger about 1.4 times NEA's six-month rent support
Why. Incidence analysis of the ledgers (29 October 2026); all figures derived and marked so.
Before. (no independence statement in the PDFs)
After. An Independence row on every Document Control page, and on the change log's: the disclaimer of website/BRIEF.md §7.6
Why. AGENTS.md requires the disclaimer on every public page; the PDFs lacked it. The Document Control page is set tighter so it still fits one page; the Executive Summary stays at 8 pages.
Before. only 4% of the 7,000 cooked-food stalls pay rent above the assessed market rent of $1,250
After. only 4% pay above their own stall's assessed market rent, whose median is about $1,250 (Dr Koh's correction by written statement, 13 Nov 2024)
Why. Dr Koh corrected the figure the same day (Official Report, SG Parliament debate 27997); each stall is compared with its own AMR.
Before. Quality tenders, the foreign hand and the cheap plate credited mainly to the PSP and Louis Chua
After. Credited even-handedly: Perera's 2021 discount cards and Chua's CHAS discounts repeated at COS 2026; Tan Yia Swam (2023), Jessica Tan and Foo Cexiang (2026) and Jamus Lim (6 Oct 2026) on a foreign hand, answered at least six times; a new 'Quality in the tender' item (Perera, Chua, Liang Eng Hwa, Mariam Jaafar, Jamus Lim; HDB's 40 of 65 tenders to other than the highest bidder)
Why. The Hansard review (29 October 2026) found cross-party proposals uncredited. Louis Chua's credits of ba2474b are kept and not repeated.
Before. (no Government answer to the academy proposals)
After. Dr Koh's reply, "we already have a similar academy in place" (the Asian Culinary Institute); the apprenticeship takes HDP entrants and adds the master's side
Why. Even-handed attribution: the Government's answer belongs beside the proposal.
Before. Melvin Yong asked whether a ballot at a fixed rental rate had been considered
After. a ballot "at a fixed rental rate targeted at cost recovery"
Why. The Official Report of 14 October 2024 (SG Parliament 27520); the phrase distinguishes it from a turnover rent. Same fix in proposals.json, whose note also wrongly called Leong's a flat rent.
Before. (no evidence that price is a poor test)
After. Of HDB's 65 eating-house tenders scored on price and quality since 2018, 40 went to someone other than the highest bidder (MND, 4 Aug 2026)
Why. Grounds the case for allocation other than by the highest bid.
Before. Private operators under government contract managing centres.
After. adds: the model came from the 2011–12 Hawker Centres Public Consultation Panel, which Grace Fu put on record on 6 March 2012
Why. Even-handed: the operator-run model has a public-consultation origin (SG Parliament 18022).
Before. (no record of earlier asks)
After. No party has asked Parliament for CPF or injury cover for hawkers; Joan Pereira (PAP) asked about leave and insurance at COS 2022 and the reply was MediShield Life and private insurance
Why. The Hansard review; Official Report of 7 March 2022 (SG Parliament 3394).
Before. (no Government statement on affordability thresholds or price data)
After. Grace Fu (7 Oct 2026): no single affordability threshold; Dr Tan See Leng (7 Oct 2026): comparative cooked-food prices across outlet types "are not available", the gap the Ledger's floor-cost and price lines fill
Why. Replies to Victor Lye (SG Parliament 34137, 34150).
Before. The 2024 debate cited from the party pages, and the Official Report for the motion and resolution only
After. The Official Report (SG Parliament mirror, debates 27944–27948, and the correction 27997) cited beside the party pages, marked as an unofficial reproduction; the 2012, 2026 and COS references added
Why. Primary record first, party copies second.
Before. Master Fellowship $3,500 a month for 12 months ($42,000), CPF $5,250, $77,250 a Fellow over one year
After. Teaching year $2,500 a month for 12 months ($30,000), then a 24-month stewardship at $1,000 a month ($24,000); CPF 12.5% on the whole stipend ($6,750); $90,750 a Fellow over three years, budgeted in the year the Fellowship starts
Why. Uncle James, 29 Oct 2026: less each month for longer, so a Fellow does not fall off a cliff when the teaching year ends
Before. Three pilot centres, Years 0–3: $18,863,142; gaps $12.76M; ladder $12.21M; corpus 133% at Year 20; keeps real value in 78% of futures
After. $19,581,112 (+$717,970; +$4,590,772 on 1.2); gaps $13.48M; ladder $12.90M; corpus 129%; 76% of futures; from Year 4 $4.13M a year
Why. Re-run of funding_model.py with the new Fellowship terms and the Regulars' List costing
Before. Heritage Watchlist lines: coordinator; verification visits and yearly check-ins at $30 stipend + $10 travel; $10 decals; price-board listing: $99,000, $107,000, $114,000 in Years 1–3
After. Regulars' List lines: coordinator; for each new listing a visit ($10 travel, no stipend), a $5 decal at print cost, a Registry page and thank-you notes ($40); a new decal when a stall moves up a tier; each centre's yearly booklet and map at all 123 centres; price board and order book: $132,750, $124,900, $121,050
Why. Uncle James, 29 Oct 2026: one way on, no way down, no re-test; the same small things for everyone listed
Before. Maintenance rate fed by the Heritage Watchlist's at-risk flow; 10% sits below the 15–25% at-risk stock
After. Fed by the at-risk count from hawkers' voluntary signals (no successor in view, stopping within three years) and public facts; a stock on a three-year horizon, 5–8% a year, so 10% is a ceiling above the declared flow
Why. The at-risk count is separate from the list and never names a stall (Uncle James, 29 Oct 2026)
Before. The Watchlist's at-risk flags, the Registry and the Hawker Council choose who is offered a Fellowship
After. A Fellowship is offered when a hawker wants to teach and the Registry's open test says the dish is fragile; declining changes nothing
Why. Going further is opt-in (Uncle James, 29 Oct 2026)
Before. The National Hawker Endowment; 'The Heritage Watchlist' as a parameter label
After. The Hawker Endowment; 'The Regulars' List'
Why. Uncle James, 29 Oct 2026: printed names (the slug national-hawker-endowment is kept)
Before. (none) the Watchlist had no tiers
After. New D.4.2, 'The Regulars' List: What a Tier Means in Plates': Bronze (the listing) always 20 named regulars; Silver 50, Gold 100, Platinum 250, or 8%, 15%, 40% of weekly plates, whichever first; a 35% healthy base (ASSUMPTION); table for every archetype and for John and Ah Kow
Why. Uncle James, 29 Oct 2026 (08:45): tiers by endorsement count, and a simulation of the share of plates from regulars a healthy stall needs
Before. Seven voting directors; NEA and NHB liaisons attend without a vote; Government holds no seat
After. Eight voting directors, one a Government director nominated by NEA or MSE, holding a casting vote and nothing more: it decides any tied decision, and the reserved matters that touch public assets (which centres host the pilots, the placement of graduates and Fellows at public stalls, anything needing NEA's or a Town Council's consent) always go to the full board; it cannot override a Hawker Council veto; every use is published within 30 days with the reasons, in the minutes and the annual report
Why. Uncle James, 29 Oct 2026: governance control must be ceded to regulators, but only through a tie-breaking vote, so that their actions are publicly accountable.
Before. Hawker Council vetoes over compensation policy, documentation consent rules, stall-placement decisions and private pilot centres
After. Vetoes plainly internal to the Seed Bank: over its own compensation policy, its documentation consent rules, the Seed Bank's own placement of its graduates and Fellows, and the board's recommendation of a private coffeeshop or food court as a pilot centre; vetoes exercised by a majority
Why. Uncle James, 29 Oct 2026: the Council's vetoes bind the Seed Bank, not public bodies.
Before. (no statement of what binds public bodies); Town Councils and operators 'come in as partners'
After. Nothing in the plan binds NEA, a Town Council or an operator; every agreement the plan needs from them is listed in §7.7 as an ask, with a new row: host the pilot centres (NEA, and the Town Council where it manages the centre, once the hawkers have voted)
Why. Uncle James, 29 Oct 2026; consents are asks, not powers.
Before. FMAS invited to name a non-voting observer at Council meetings
After. FMAS names the holder of a tenth Council seat, with a full vote
Why. Uncle James, 29 Oct 2026: FMAS gets one voting seat on the Hawker Council, not an observer.
Before. A recognised voice on tenancy and operator-contract terms · No bargaining body
After. A recognised voice: FMAS and the centre hawkers' associations, recognised as the trade's bargaining body, not a new one · No bargaining body with a mandate
Why. Uncle James, 29 Oct 2026: the bargaining body means FMAS and the centre associations.
Before. the notice is deemed approved unless NEA objects within two working days
After. as we ask of NEA, the notice is deemed approved unless NEA objects within two working days
Why. Uncle James, 29 Oct 2026: label what is asked of NEA as an ask.
Before. an ombudsman reporting to MSF
After. an independent ombudsman, appointed by the board and reporting publicly each year
Why. Uncle James, 29 Oct 2026.
Before. Launch thresholds (40% willing, 100 signatures, a Council majority)
After. The same thresholds, said to be the Seed Bank's own test; the centre's hawkers' vote and NEA's or the Town Council's agreement are the conditions outside it (§7.7)
Why. Uncle James, 29 Oct 2026.
Before. Fellowship: $3,500 a month for 12 months ($42,000), CPF $5,250, $77,250 a Fellow over one year
After. A teaching year at $2,500 a month ($30,000), then a 24-month stewardship at $1,000 a month ($24,000) for light duties (tastings, locum tests, two visits a month to the apprentice's stall, one public session a year); completion grant and health cover unchanged; CPF 12.5% on the whole stipend ($6,750); $90,750 a Fellow over three years, each cohort's three years budgeted in the year it starts; intensive track compresses the teaching year then the same stewardship; Emeritus pathway follows the stewardship
Why. Uncle James, 29 Oct 2026: a Fellow should not fall off a cliff when the teaching year ends; less each month, for longer.
Before. The Heritage Watchlist: four qualifying tests, a verification visit, a yearly four-question check-in with $30 + $10 visit stipends, triage on two signals (65+, no successor, tenancy ending, last stalls), decal 'On the Heritage Watchlist'
After. The Regulars' List: a compliment from diners ('the regulars of this centre have asked that this dish be kept'); listed at Bronze on 20 endorsements by weekly regulars (one each, verified by a mobile number, never bought) or an association nomination; Silver, Gold and Platinum at higher counts or plate shares, whichever first; tiers only go up; leaves only by the hawker's choice or closing; no re-test; the same small things for all (Registry page, decal with tier and year, price board and order book, booklet and map, thank-you note); no stipends or decal fees; Fellowships opt-in, offered when the hawker wants to teach and the Registry's open test finds the dish fragile, never gated by tier; the at-risk count kept apart, from two voluntary signals (no successor in view, intention to stop within three years) and public facts, published only as national and per-centre totals
Why. Uncle James, 29 Oct 2026 (08:30 and 08:45): simplify the working assumptions, make the list a compliment that does not hurt a stall that stays on it, and tier it by endorsements.
Before. Seven selection tests incl. footfall and a mix of managers; operator-run centres suggested as pilot or comparison ('none of the three is operator-run')
After. One ranking, the expected count of heritage-worthy stalls (Year 0 proxy, then the Regulars' List's own count), and four conditions: a government centre (private by exception), every community's cooking with Geylang Serai by default, no major works in the first year, and consent; all 123 centres eligible, none favoured for who runs it; Tekka third by the community condition, not the count, said plainly; operator-run pilot and comparison sentences removed; the evaluator may use MSE's published SEHC figures (one clause, §6.4)
Why. Uncle James, 29 Oct 2026: remove the operator-run pilot requirement; prioritise by the likely concentration of heritage-worthy stalls.
Before. Hawkers Succession Scheme 2.0
After. The Hawkers Succession Scheme, Extended (heading); 'extending the Hawkers Succession Scheme' elsewhere; section id and proposal id hss-2-0 kept
Why. Uncle James, 29 Oct 2026.
Before. the National Hawker Endowment
After. the Hawker Endowment (proposal id and slug national-hawker-endowment kept)
Why. Uncle James, 29 Oct 2026.
Before. Makan tranche: a permanent tranche on live CDC rails, ≈$66M a year
After. Presented as a hawker tag on voucher envelopes the State already funds: about 12% of the hawker-and-heartland half of CDC and SG60 voucher spending (more than $2.1bn over four years, about $525M a year; MCCY, 3 Feb 2026; the share derived); $66M kept as the indicative size; carve-out or top-up left to the Budget
Why. Uncle James, 29 Oct 2026.
Before. Family Name Royalty: 2.5% of gross, in perpetuity ('forever', 'a permanent stake')
After. 2.5% of gross for ten years at most from its first month (about $55,000 for a $220,000 stall), with a buy-out at any time at a multiple the Seed Bank publishes in advance; why ten years said plainly
Why. Uncle James, 29 Oct 2026. The stress test is unchanged: the cap never binds within its ten-year horizon (numbers_A3 §12).
Before. Heritage Watchlist (§3.6) — … verification visits / check-ins …; Master Fellowships — N × $77,250
After. Regulars' List (§3.6) — coordinator; new listings, with decals; Registry pages, booklets and maps; its listing; Master Fellowships — N × $90,750, three years each
Why. Ledger rows follow the new terms; amounts from the model (numbers_A3.md).
Before. $18,863,142 over Years 0–3 (+26% on 1.2); 72.1%; Watchlist $99,000 / $107,000 / $114,000; Lighthouses alone $15,568,579; add-ons $5,304,561 and $4,191,829; gaps $12.8M; 78 of 100 futures
After. $19,581,112 (+30.6% on 1.2: the Fellowship stewardship and CPF about $0.91M, the Regulars' List about $0.06M besides the apprentices and the relief day); 72.8%; Regulars' List $132,750 / $124,900 / $121,050; Lighthouses alone $16,152,899; add-ons $5,542,161 and $4,384,879; gaps about $13.5M; 76 of 100 futures; $1.08 public per private dollar; retirement-match shortfalls $11.2M / $28.9M
Why. The model re-run of 29 Oct 2026 (numbers_A3.md).
Before. Nineteen Plate Tracker indicators: Singapore publishes 3 regularly, 10 in part and 6 not at all
After. Twenty: a new row, 'Stalls at risk, and the Regulars' List' (the voluntary at-risk count as national and per-centre totals, never naming a stall, and the listed stalls by tier), missing today; 3 published, 10 in part, 7 not at all
Why. The at-risk count of §3.6 is published only as an aggregate; the Tracker carries it so that the count and the table stay in step
Before. (no statement of what the Council's tenth seat and the Government director cost)
After. The line for Hawker Council fees and directors' stall-visit vouchers is a lump sum; the Council's tenth seat, for FMAS, and the board's Government director sit inside it (§3.8)
Why. The Council and the board each gained a seat in this release; the line is not per member, so it is not re-costed
Before. about half ($10.22M) goes to lines that only the 44 heritage-worthy stalls receive … more than a quarter runs the institution
After. about half ($10.17M) … about a third runs the institution
Why. Re-derived with the incidence method on the new model run: the Fellowship rise is $594,000 gross, since the 10% contingency is counted with the institution, and the institution's share is 32.5%
Before. From Year 4 … about $4.13M a year … against about $5.31M from the draw, the operating match and continuing gifts
After. … against about $5.32M
Why. The Year 4 draw ($1.52M), the operating match ($0.50M) and continuing gifts ($3.29M) come to $5,316,908; the figure had been truncated, not rounded, since 1.2
Before. Web pages at …/hawkers-succession-scheme-2-0-add-the-engine-to-the-rail/, …/the-heritage-watchlist-and-the-endangered-dish-registry/ and …/funding-stacks-and-the-national-hawker-endowment/
After. The pages take their new headings' slugs; every 1.2, 1.1 and earlier route to them is sent to the new page by a single 301
Why. The three headings were renamed in this release (the Scheme extended, the Regulars' List, the Hawker Endowment)
1.2 28 October 2026 8 changes · shared version
A restructure for readability, with no change to any proposal, figure or claim. The three papers become a library of standalone documents sharing one version: the Executive Summary; the Plan, the case and every proposal; and six companion volumes – Who Already Helps (A), the Budget Book (B), Hawkers and Social Protection (C), Follow the Plate (D), Succession Economics (E) and Method, Levers and Data (F) – each opening with who should read it, its bottom line in five points, how it connects and the terms it uses. The Main Paper and the Technical Paper end; every section moves, with a new id, and every cross-reference names the document it points to. Under the versioning rule Uncle James set on 28 October 2026, a restructure by the initiator alone is a minor version; a new edition, with contributors or reviewers beyond him, is a major one.
Approved by Uncle James · tag v1.2 · release commit 298c540
Before. One document in three papers: the Executive Summary (8 pp), the Main Paper (86 pp) and the Technical Paper, Under the Hood (73 pp), with the Change Log beside them.
After. One document as a library of nine PDFs: 1 Executive Summary; 2 The Plan; A Who Already Helps; B The Budget Book; C Hawkers and Social Protection; D Follow the Plate; E Succession Economics; F Method, Levers and Data; 0 the Change Log. The Main Paper and Technical Paper PDFs are removed; the repository keeps one copy of each document.
Why. Uncle James (28 Oct 2026): "separate each appendix chunk into standalone PDFs, so that the download page by itself makes sense, with each major appendix block being a document with standalone reading viability. I'm keen to significantly uplift readability regardless of where they start." Moving the detail into one appendix would only have made the appendix too large.
Before. The Main Paper reprinted the Executive Summary after its contents; Book One and Book Two and each of the seven parts opened on a divider page of their own; §1.3 printed §1.3.1–§1.3.2 in full; Part VII printed the four ledgers, the endowment's funding, the ask and the downside in full; the references closed the paper.
After. The Plan opens with its cover, The Plan at a Glance (the former Plan on One Page) and a contents page that points to the Executive Summary PDF and the six volumes. Each part's divider is merged with its 'In one minute' box and first section; the Book dividers go. §1.3 keeps Government's rails and the findings in about two pages, with Fig. ES.1, and points to Who Already Helps. Part VII keeps §7.1's yearly totals, the operating and programme split, each centre's own budget and the add-ons; §7.6's endowment in short; §7.7 and §7.8 in short form, with an 'In full' line to the Budget Book; §7.9 in full. A short sources note replaces the references, now in F.5.
Why. The Plan should read as the case and the proposals, without the ledgers and the review that only some readers need.
Before. §1.3.1 Beyond Government: Who Else Already Helps and §1.3.2 What Only This Plan Does, in the Main Paper.
After. Who Already Helps: A.1 Standing on What Works (§1.3's rails, reprinted so the volume reads on its own), A.2 (§1.3.1) and A.3 (§1.3.2).
Why. A partner or charity can read the whole review in one short document.
Before. The ledgers, the endowment, the ask and the downside in Main Paper Part VII; Follow the Gift in Technical Paper T1.4, with the heritage-stall method in T1.4.6; the gates and funding levers in T2.7–T2.8.
After. The Budget Book: B.1 Reading the Ledgers (§7.1, with each centre's budget and the add-ons), B.2–B.5 the ledgers for Years 0–3 (§7.2–§7.5), B.6 the endowment (§7.6, with B.6.1 and B.6.2), B.7 the ask itemised in full (§7.7), B.8 downside mode and sunset in full (§7.8), B.9 Follow the Gift (T1.4; the heritage-stall method is B.9.6), B.10 and B.11 the levers (T2.7, T2.8).
Why. A funder can read every line of the money, and the model that tests it, in one document.
Before. Social protection compared in Technical Paper T1.1 (T1.1.1–T1.1.7); the Protection Floor's levers in T2.6, with the retirement match costed in T2.6.1.
After. Hawkers and Social Protection: C.1–C.6 (T1.1.1–T1.1.6), C.7 Whoever the Landlord (T1.1.7), C.8 the levers on the Protection Floor (T2.6) and C.8.1 the retirement match (T2.6.1). §2.11, the Floor itself, stays in the Plan.
Why. Labour and CPF policy readers get the comparison and its levers together.
Before. —
After. Each companion volume opens with a one-page opener in Uncle James's voice, drawn from its body: 'Read this if you…', 'The bottom line' in five points, 'How this connects' (where it comes from, what to read with it, what is in it) and the terms it uses; and closes with a short sources section and its own Document Control page. The text is kept in src/volumes/. No proposal, figure or claim is added.
Why. Every volume must read on its own, wherever a reader starts.
Before. Cross-references used the three papers' ids (§1.3.1, T1.2.4.1, T2.5) and names (Main Paper, Technical Paper).
After. Every cross-reference uses the library's ids and names the document it points to when that is another one ('the Budget Book, B.7'; 'Follow the Plate, D.4.1'); references within a document print the id alone. The Executive Summary's connective lines name the Plan and the companion volumes; its page numbers are the Plan's. Old ids map to their new homes in src/layers/refs.py and src/layers/legacy-map.json, and every old web route redirects.
Why. No reference may send a reader to a document that no longer exists.
Before. A new major version is a new edition; the papers' structure or core changes is a major version.
After. A new major version is a new edition, made with contributors or reviewers beyond the initiator; a restructure by the initiator alone is a minor version. Document Control lists the whole library.
Why. Uncle James (28 Oct 2026): "keep it as v1.2; v2 should involve others beyond me." 1.2 is the first application of the rule.
1.1 27 October 2026 43 changes · shared version
First revision of 1.0. A review of who already helps hawkers beyond Government (hawkers' own associations, the labour movement, the five Mayors and CDCs, the People's Association and grassroots, Town Councils, charities, social enterprises, companies and researchers) and of overseas precedents, so that the plan builds on them and duplicates none. The Executive Summary gains a comparison of hawkers' protection with four other kinds of worker and a summary of the review; the retirement match now applies under 55, beside the existing scheme from 55. The pilot widens from one Lighthouse Centre to two Lighthouses from Year 1, Chinatown Complex and Geylang Serai, and Tekka from Year 2, five at most before the gate, and the programme is re-costed centre by centre from each centre's heritage-worthy stalls and size. Hawker-led Buying Circles (§2.12) and two asks of Government, a table-cleaning fee re-priced for tray return at the pilot centres and a named deputy for an hour a day, join the package, and the prep co-operative is dropped from the model as well. Earlier proposals are attributed even-handedly, each proposer named in the same parliamentary terms whatever the party, with what the plan agrees and disagrees with. A light first tier, the Heritage Watchlist (§3.6), opens at every government hawker centre from Year 1 and gives a yearly count of the stalls at risk; from Year 4 each centre funds new Fellowships at a maintenance rate fed by that count.
Before 1.2 the plan was three papers, the shared ancestors of both stalls; this entry is in both logs.
Approved by Uncle James · tag v1.1 · release commit b4fe4b7
Before. ES.1–ES.6.
After. New ES.3 The Thinnest Floor, and No Business to Sell (stallholder, assistant, platform worker, lower-wage employee, PMET) and ES.4 Who Already Helps, and Where This Plan Fits; the later sections renumbered ES.5–ES.8.
Why. Uncle James asked for the gaps hawkers face to be shown against gig, blue- and white-collar workers, and for the review of existing help to be summarised in the Executive Summary.
Before. §1.3 covered Government's rails only; §1.2 described FMAS as 'minority coverage, no leverage'.
After. New §1.3.1 Beyond Government: Who Else Already Helps (a table of 15 groups of actors and where the plan joins, defers or fills) and §1.3.2 What Only This Plan Does — and Where It Builds on Others (gaps no one fills; what is not new; overseas precedents). §1.2 describes FMAS fairly: wide reach, no welfare fund and no decision rights of its own.
Why. Make sure the plan has no blind spots and duplicates nothing already done by non-profits, charities, the Mayors and the People's Association.
Before. The retirement match applied at any age: base up to about $8.3M a year, stretch about $14.1M; shortfall over Years 4–10 $31.2M base, $55.5M stretch with HDB ($65.4M without); the endowment would need about $240M to carry the base.
After. The match applies to stallholders under 55; from 55 the Matched Retirement Savings Scheme already matches top-ups (up to $2,000 a year), and the plan helps hawkers claim it. Base at most about $4.1M a year, stretch about $8.0M (ceilings: fewer than half of stallholders are under 55); shortfall over Years 4–10 $6.8M base, $24.8M stretch with HDB ($29.8M without); about $120M of endowment would carry the base. Package programme cost about $80M (was $85.9M).
Why. The existing scheme already matches top-ups from 55; a second match would pay twice.
Before. The proposals did not name the existing actors working in the same space.
After. §2.5 asks that CDC targeted meal credits be spendable at hawker stalls, works through Project Belanja! and grassroots meal cards, and keeps the Affordability Compact funded, never contractual; §2.4 names the Foochow coffeeshop association; §2.7 recruits through Jobs Nearby @ CDC; §2.8 puts local pipelines first; §2.9 extends the FMAS–NTUC group cover; §2.10 adds cost data to price data others already publish.
Why. Build on what exists rather than beside it.
Before. These sections did not say what was already done or proposed.
After. §2.2 answers the 2024 parliamentary ballot proposal and the Government's reply; the Seed Bank's archive and apprenticeships acknowledge and partner existing documentation and training; the Hawker Council works with FMAS (a non-voting FMAS observer) and centre associations; the Fellowship and Taste Council build on the Hawkers Succession Scheme's stipend and tasting panel; the Heritage Plate is set apart from existing awards and labels; the Lighthouse is an NEA-managed centre; the Seed Bank registers under Arts and Heritage for the Cultural Matching Fund; the CDC Vouchers Donation Scheme is a giving route.
Why. Uniqueness and no duplication; each change follows the review in §1.3.1–§1.3.2.
Before. The food-stall assistant wage floor shown as an employee protection; the 2023 median wage ($5,070); the platform CPF rate without its age band; the Budget 2026 CPF Top-Up without its eligibility; about 6,900 cooked-food stalls attributed to NEA.
After. No wage floor binds a hawker stall; the 2025 median ($5,775, about 642 plates a day); platform and senior employee CPF rates by age; the Top-Up's eligibility and where it is paid; Jobseeker Support's $5,000 income cap; 6,900 marked as our estimate; a note splitting the employee column into the hawker's assistant, lower-wage employees and PMETs.
Why. Fact-check against 2025–26 official sources.
Before. —
After. 92 new sources for the review.
Why. Every new claim is sourced.
Before. ES.4 summarised the review in a list.
After. Fig. ES.1, a map of sixteen needs across a hawker's working life: who helps today (Government agencies, hawkers' own bodies, the Mayors and CDCs, the PA and grassroots, charities and others), what the plan does about each (fills a gap, builds on or joins), with its section, and who pays for the plan's part (the State, gifts and the endowment, or the trade itself). Drawn by src/layers/support_map.py.
Why. Uncle James asked for a visual map of existing support against the gaps the plan fills, with who pays as a further dimension.
Before. Fig. ES.1 showed hawkers' own bodies recognising masters at scale, and the plan building on it; ES.4 said they "recognise its masters".
After. Hawkers' own bodies recognise masters partly, and the plan reinforces that recognition (§4.2). ES.4: they "honour a few of its masters each year – with an award, not a wage", and the gaps include paying a master a year's wage to teach. §1.3.1 notes that the FMAS awards carry no published prize money, stipend or teaching role, that the Fellowship adds a year's pay to teach, and that FMAS's standing as the trade's voice is wide but not uncontested.
Why. Uncle James questioned whether hawkers' own bodies recognise masters sufficiently and meaningfully. They hold one ceremony a year with no published cash, so "at scale" overstated it and contradicted §4.2.
Before. Fig. ES.1 marked nine needs "Fills a gap", seven of them where some help already exists.
After. "Fills a gap" is kept for the two needs no one meets today (shared prep kitchens; repairs and exhausts). Retirement savings keeps "Fills, under 55". A new mark, "Reinforces it", shows the seven needs where help exists but is thin and the plan adds what it lacks: income once in a stall, days off, injury and illness cover, paying for goodwill, recipes kept to cook from, recognising masters and money that lasts. The legend gains a line for the plan's marks, and ES.4 says which rows are empty and which the plan reinforces. support_map.py now checks that "Fills a gap" appears only on rows where nothing was found.
Why. Uncle James pointed out that the plan reinforces areas the map shows as partly covered, and asked that the map say so.
Before. The 1.1 text drafted to a style sheet taken from the papers alone.
After. The same facts, re-voiced to Uncle James's own writing (his blog, essays and company updates).
Why. Uncle James asked that his writing style be studied from his own published writing.
Before. The plan proved everything at one Lighthouse Centre in Years 1–2. Year 3 was gate-dependent expansion: a second co-op site and the Security Package at a second centre, $600K in all. Expansion beyond the Lighthouse waited until the evaluator certified five gate criteria at month 20 or later (Fellowships ≥70%; first cohort landed, ≥80% operating at month 12; trust ≥55% and rising; three Succession Mortgages; the second endowment close). The rent-band pilot and the 50 Apprentice Work Passes were at the Lighthouse and its cluster.
After. One Lighthouse Centre opens in Year 1, as before. In Year 2 the plan widens to three pilot centres, five at most, alongside the Lighthouse. Each later centre passes its own §6.2 veto with its own hawkers willing, gets its trust deposits (the Security Package, listening and co-design) in the six months before it opens, in Year 1, and runs the full loop: Master Fellowships, apprentices at the masters' own stalls, a Relief-Corps bench, the Security Package, the rent-band and ballot pilot and evaluation. The Taste Council, the repository and the core institution stay national and shared. No further than five until the evaluator certifies the five gate criteria, now measured across all the pilot centres at month 43 or later after the Lighthouse opens (Fellowships ≥70% with no centre below 50%; every centre's first cohort landed, ≥80% operating at month 12; trust ≥55% and rising at every centre; at least six Succession Mortgages at two centres or more; the final close, with each further centre's first two years funded). The gate governs going beyond five, from Year 4, not a Year 3 second site. The rent-band pilot and the 50 passes run across the pilot centres and their clusters.
Why. Uncle James (27 Oct 2026): "We need more data points than just one Lighthouse Centre. Not every hawker is going to be willing to try our schemes out."
Before. Four-year programme $9,125,460 (Year 0 $711,700; Year 1 $2,358,020; Year 2 $2,467,420; Year 3 $3,588,320, of which $600K gate-dependent): two co-ops, 18 Master Fellowships, 36 apprenticeships, 13 landings. From Year 4, $2,988,320 a year, indexed. Gaps of about $3.1M across Years 1–3. No later year short; the corpus kept its real value in about 19 futures in 20. Retirement-match shortfall over Years 4–10 $6.8M base, $24.8M stretch with HDB ($29.8M without). Downside: Lighthouse-only Year 3 roughly $2.9M. About $1.61 of public money for every private dollar in the campaign. First-loss budget $410,000.
After. At three centres, the central plan, $12,521,650 (Year 0 $526,900; Year 1 $2,215,990; Year 2 $4,272,430; Year 3 $5,506,330), after the 1.1 removals below: 38 Master Fellowships, 76 apprenticeships, 23 landings. Each further centre, the fourth and the fifth, is a stated add-on of $2,570,140 ($196,900, $994,770, $1,378,470); five centres $17,661,930. From Year 4, $5,506,330 a year, indexed, and $1,378,470 more for each further centre. Gaps of about $6.7M across Years 1–3. The programme spends every continuing gift and is $1.7M short over Years 4–10 ($10.9M over Years 4–20); the corpus keeps its real value in about half of futures (52 in 100). Retirement-match shortfall $15.7M base, $33.6M stretch with HDB ($38.6M without). Downside: the Lighthouse alone, $2,663,590 in Year 3. About $1.37 of public money for every private dollar. First-loss budget $570,000.
Why. Uncle James: "Let the budget rise, and say so." Each later pilot centre's lines are the Lighthouse's own, one year behind, with its deposits in Year 1; a shared pilot-centre coordinator, per-centre evaluator, Taste Council and events lines, and the 10% contingency. Revenue is scaled per centre, at what the Lighthouse earned at the same age. Every new assumption is labelled ASSUMPTION in the funding model's notes (T1.4).
Before. The 5,000 futures varied whether the gates opened for Year 3's $600K expansion (70%).
After. They vary whether each of the two later pilot centres passes its hawkers' vote (80% each; a centre voted down keeps its Year 1 deposits and costs nothing after). Four and five centres, and the Lighthouse alone, are run as scenarios.
Why. The gate no longer governs any Year 3 spend; a centre's own veto is the uncertainty the plan cannot control.
Before. The Hawker Security Package gave three deposits at each pilot centre before any ask: group critical-illness cover for stallholders 55+, an exhaust and ducting deep-clean blitz across ~120 stalls ($78,000), and a same-week equipment repair fund ($90,000), with yearly renewals of insurance and repairs ($105,000–$145,000 a centre). The Package was billed as the successor to the Hawkers' Productivity Grant. Fig. ES.1 showed "Repairs and exhausts" as a gap the plan fills.
After. The Package is cover only: the group critical-illness pool for stallholders 55+ ($52,500 a centre a year, about 70 at about $750), and the FMAS–NTUC group-cover route widened to critical illness and to stallholders. Exhausts, ducting and stall repairs are the landlord's (NEA or the Town Council) and the hawker's own business, and the papers say so. Fig. ES.1 keeps the need, marked "Left to landlords", with no section and no payer.
Why. Uncle James (27 Oct 2026): "Why are you budgeting on infrastructural items of the hawker centre? That's not our programme objectives." The savings are banked, not re-spent.
Before. §2.6 Prep Co-operatives — Twin Certified Lines: a hawker-owned co-op kitchen at each pilot centre with halal and non-halal lines (fit-out $420,000 at the Lighthouse, $380,000 at each later centre; operations $60,000–$85,000 a year; member fees $150–$300 a month; a co-op manager in the core team), claiming a 14–16-hour day compressed towards 10–12. Programme-run heritage stalls staffed by apprentices (2, 3, then 5 at the Lighthouse; 2 then 3 at each later centre; setups, rent subsidy and stall net). The incumbent package in Follow the Plate included the co-op.
After. §2.6 Learning at the Master's Side – Nothing to Build: apprentices learn at the master's own stall as registered stall assistants (building on the Hawkers Succession Scheme), each record is proved on the master's relief day, and blind tastings and retired masters' documentation book partner kitchens by the session; graduates land through the rent-band ballot or the Incubation Stall Programme. The only costs are a one-off $3,000 second-station grant at each new Fellow's stall and about 24 kitchen sessions a centre a year at $300 (both ASSUMPTIONS). No co-op, no programme-run heritage stalls, no co-op manager; the plan no longer claims a shorter prep day. Follow the Plate keeps the co-op only as a tested variation, outside the package: with the package a tendered stall stays above the assistant benchmark up to about $1,660 (was $2,280), and 41–55% of tendered stalls fall below it across $1,200–1,800 (was 28–41%).
Why. Uncle James (27–28 Oct 2026): a working hawker will not leave his stall to prep elsewhere and come back; shared prep only makes sense for cooking recorded dishes to spec, which is training. He chose an asset-light design that takes no stalls (approved 00:15 SGT, 28 Oct).
Before. Each year's ledger was one list of lines, at gross cost, with a 10% contingency, programme revenue and a net requirement.
After. Each year's ledger has two parts: A, the operating budget (the institution's burn: core team, pilot-centre staff and liaison officers, interim lead and research assistants, office, legal, audit and insurance, communications, the evaluator, Hawker Council fees and the shared national platforms), and B, the programme budget (grants and expenses that reach hawkers), shown per centre for the Lighthouse and pilot centres 2 and 3, each with its contingency, revenue deducted from B; the net requirement is A + B. Each year prints the share reaching hawkers directly (66.0% over Years 0–3) and the operating burn a month ($88,800 on average). §7.1 adds a summary of the split and a budget for one pilot centre. The funding model carries an operating and a per-centre programme line (funding.js in parity).
Why. Uncle James asked to split the programme staff budget from the hawker programme grants and expenses, "like a tech startup".
Before. Follow the Plate kept the prep co-operative as a tested variation outside the package ($225 a month, ingredients 2% and utilities 10% cheaper, the day from 13 to 11 hours): T1.2.7 scored it (leverage 0.22, 3.57 with time), T1.2.5.2 and T2.4 described it, and the simulator offered it.
After. The co-op is gone from the model (plate_model.py and its browser port, in parity), the simulator, the site's Follow the Plate page and the Technical Paper. The Main Paper keeps only §2.6's account of why it was dropped.
Why. Uncle James (28 Oct 2026): drop the prep co-op completely, including the tested variation.
Before. No proposal helped hawkers buy together; the papers called buying together a gap.
After. New §2.12, The Buying Club – Buy Together, Own Nothing: members buy ingredients on negotiated wholesale contracts, delivered to the stall, through software with no warehouse; one power contract on the open electricity market, with gas through the same club; delivery at a 15% group commission or through a commission-free channel building on Hawkers United – Dabao 2020. ASSUMPTIONS: ingredients 3% and utilities 10% cheaper, commission 15% (was 22%). Self-funding with no member fee: a 1% supplier rebate on about $40,000 a member a year (ASSUMPTION) against a procurement and partnerships lead (1 FTE, $78,000) and ordering software ($30,000, then $18,000 a year) in the operating budget; it pays for itself from about 265 members, and at three centres the operating budget carries $216,000 over four years. It concedes operators' bulk buying, little used (MSE, 10 Jan 2024), and builds on it without a new agency. In the package.
Why. Uncle James (28 Oct 2026): add asset-light levers, all of them, to the package; costs in the ledgers, the role in the operating budget.
Before. §2.2 treated rent alone as the occupancy cost; the table-cleaning fee (2025 median $520 a month) was not discussed.
After. §2.2 adds the fee beside the rent: an ask of NEA and operators to re-price the table-cleaning fee for the work left since tray return became the rule in 2021. ASSUMPTION: $200 a month less, at NEA and operator-run centres. The model books it as income the fee collector gives up (landlord share), an upper bound: about $16.6M a year across 6,900 stalls, $16.0M of it the State's and $0.5M the operators', counted as programme cost; shown in T1.2.2.2, T1.2.5, T1.2.7.1 and T1.2.7.4, and as a row in §7.7.
Why. Uncle James (28 Oct 2026): the fee was set before tray return; ask Government and operators to re-price it, and show who bears it.
Before. The relief notice was the only change asked of the personal-operation rule.
After. §2.7 adds a named deputy, an NEA rule change: one named, registered assistant may run the stall for up to an hour a day, logged like the relief notice, while the stallholder stays on the roll. It concedes that the personal-operation rule exists to stop subletting and sets out the safeguards. ASSUMPTION: the hawker works 12 of his 13 hours, at no extra wage. Zero fiscal cost; a row in §7.7. In the package.
Why. Uncle James (28 Oct 2026): a named deputy, an ask of Government, with safeguards so it cannot become subletting.
Before. Package (price freedom, Makan Vouchers, Relief Corps, Protection Floor): a typical tendered stall above the $11.64 assistant benchmark up to about $1,660 of rent; 41–55% of tendered stalls below it across $1,200–1,800 with the package (62–75% today); $12.28 an hour with the package; national package programme cost about $77.4M a year, leverage 0.90 (cash 0.79).
After. Package with the buying club, the fee reset and the named deputy: above the benchmark up to about $2,590 (about $2,300 without the deputy); 22–34% below it across $1,200–1,800 (62–75% today, unchanged); $15.43 an hour; national package programme cost about $93.4M a year with the fee reset counted in full, leverage 1.12 (cash 1.01). Each lever alone on the first four: ingredients $1,900, utilities $1,760, fee $1,860, delivery $1,760, deputy $1,950. The successor's protection still binds from about $1,920 and a stall still loses money only above about $4,500.
Why. Recomputed with the new package.
Before. $12,521,650 over Years 0–3 at three centres (Y1 $2,215,990, Y2 $4,272,430, Y3 $5,506,330); operating $4,263,160, 66.0% reaching hawkers, burn $88,800 a month; each further centre $2,570,140; Lighthouse alone $7,123,970 ($2,663,590 in Year 3); bridge grants $6.7M; from Year 4 about $0.24M a year short ($1.7M over Years 4–10).
After. $12,737,650 (Y0 $526,900, Y1 $2,322,790, Y2 $4,336,030, Y3 $5,551,930); operating $4,479,160, programme unchanged at $8,258,490, 64.8% reaching hawkers, burn $93,300 a month; each further centre $2,540,140; Lighthouse alone $7,399,970 ($2,745,190 in Year 3); bridge grants $6.9M ($1.6M for the Lighthouse alone); from Year 4 about $0.29M a year short ($2.0M over Years 4–10); about $1.36 of public money for every private dollar in the campaign.
Why. The buying club's lead, software and supplier rebates in part A of each ledger.
Before. Fig. ES.1 mapped fifteen needs.
After. Fig. ES.1 maps sixteen: 'Buying ingredients and power' under Running the stall, marked as partly helped by others (some centre operators offer bulk buying, little used: MSE, 10 Jan 2024), which the plan reinforces (§2.12), paid for by the trade. Repairs and exhausts remain the only empty row, left to landlords.
Why. Uncle James (28 Oct 2026) asked for the row; the sources show some help exists today, so it is not marked 'nothing found'.
Before. 1.0 did not mention the earlier proposals. The first 1.1 drafts named the party behind the motion of 13 November 2024 but the Member who asked the question of 14 October 2024 without his, and said the plan 'takes the idea' of centralised procurement; a later draft named no party or Member and deleted the procurement mention. §1.3.2 still spoke of a one-centre pilot.
After. Even-handed attribution. Every proposer is named in the same parliamentary terms, with no adjectives: "Leong Mun Wai, a Non-Constituency MP of the Progress Singapore Party" moved the motion; "Melvin Yong, the People's Action Party MP for Radin Mas" asked the question; "Hazel Poa, a Non-Constituency MP of the Progress Singapore Party" proposed centralised procurement through a new agency; the House passed the motion as amended by "Edward Chia, the People's Action Party MP for Holland–Bukit Timah". The Government's answers stay. The plan says what it agrees with and what it does not, and judges the idea, not the proposer: "I agree that stalls should be allocated by ballot rather than by the highest bid. I do not agree with a flat rent" (§1.3.2, §2.2); "I share the aim of cheaper ingredients for hawkers. I do not agree with a single national buyer, even an optional one: it would narrow the competition among Singapore's suppliers, not widen it" (§1.3.2, §2.12). "$500 or 3% of Gross Turnover, whichever is higher" and "a sixth category, namely heritage foods" are quoted from the mover's own speech and marked as his words. References: the two speeches restored, marked as published by the party and cited only for the speakers' own words, beside MSE's response to the motion and its oral reply of 14 Oct 2024; the sources data gains both speeches and the oral reply. §1.3.2 now says three pilot centres, and the §7.7 rent-bands row says the plan "keeps the ballot proposed in Parliament in 2024, not its flat rent".
Why. Uncle James (28 Oct 2026, 07:34): "You can mention centralized procurement and you can note the ballot motion and earlier parliamentary questions. Don't avoid it. I just don't want us to be seen to be endorsing ideas they originate as if we agree with them. Make it clear what we agree with or not, but don't show partiality."
Before. The model note for the Lighthouse-alone operating budget said: "Years 0–2 are the Lighthouse's ledgers as first costed, less the exhaust blitz, the repair fund and the repairs in each renewal."
After. The stale sentence is deleted; the ledgers no longer carry hawker-centre infrastructure, so there is nothing to subtract. The figures are unchanged.
Why. Uncle James (28 Oct 2026) asked for the stale note to go.
Before. §2.12, The Buying Club – Buy Together, Own Nothing: a members' club run in software by the Seed Bank, negotiating wholesale contracts and one power contract, self-funding through a 1% supplier rebate against a procurement lead (1 FTE, $78,000) and ordering software; $216,000 over four years at three centres.
After. §2.12, Buying Circles – Hawkers Bargain Together. The Seed Bank never buys, stocks, sets prices or picks suppliers, and never takes title to goods or money; every hawker keeps his own supplier and recipe. Through their own hawkers' association the stalls at a centre form small circles (rice, oil, poultry, gas), voluntary, with no exclusivity, free to leave any week; a circle lead, usually a hawker, receives a small honorarium (ASSUMPTION: $100 a month). Circles post their pooled weekly volume on an open price board and suppliers compete; operators' bulk-purchase offers join the board. Power by group switch on the open electricity market, gas by circle, delivery rates negotiated per centre or commission-free channels. The Seed Bank provides the board (on the Plate Ledger's data), group-order templates and one part-time partnerships role (0.5 FTE, $39,000 a year); the full procurement FTE, the software and the supplier rebates go. Savings unchanged and marked ASSUMPTION (ingredients 3%, utilities 10%, commission 22% to 15%). No member fee; the circles do not pay for themselves at three centres, and the operating budget carries $185,460 over four years.
Why. Uncle James (28 Oct 2026): "We can't be over-centralising procurement into a single national agency; that's akin to nationalising hawker procurement, which can be sensitive and sends wrong signals to supplier entrepreneurs."
Before. The table-cleaning fee reset was asked for at every NEA and operator-run centre and booked nationally, up to about $16.0M a year to the State and $0.5M to operators, inside the national package cost of about $93.4M a year; ES.4 counted four places where the State is asked for money.
After. The reset is asked of NEA at the pilot centres only, three, five at most, so the evaluator shows its effect before anyone prices it nationally. ASSUMPTION: a pilot centre has about 120 stalls, like the Lighthouse, each paying the fee: up to about $864,000 a year at three centres ($288,000 a centre), $1.44M at five, an upper bound. The national figure (about $16.6M a year, $16.0M the State's) is shown apart as a later decision, outside the package total and the plan. A pilot-centre stall still gets the reset, so the package's per-stall effect and the break points are unchanged. ES.4 now counts five places: four Budget lines and the fee income given up at the pilot centres.
Why. Uncle James (28 Oct 2026): ask for the reset at the pilot centres first; present the national figure only as an upper bound for a later decision.
Before. $12,737,650 over Years 0–3 at three centres (Y1 $2,322,790, Y2 $4,336,030, Y3 $5,551,930); operating $4,479,160, 64.8% reaching hawkers, burn $93,300 a month; each further centre $2,540,140; Lighthouse alone $7,399,970 ($2,745,190 in Year 3, gaps $1.6M); from Year 4 about $0.29M a year short ($2.0M over Years 4–10); national package programme cost about $93.4M a year, leverage 1.12 (cash 1.01).
After. $12,707,110 (Y0 $526,900, Y1 $2,275,170, Y2 $4,335,570, Y3 $5,569,470); operating $4,448,620, programme unchanged at $8,258,490, 65.0% reaching hawkers, burn $92,700 a month; each further centre $2,580,700 (82% reaching hawkers); Lighthouse alone $7,288,310 ($2,716,170 in Year 3, gaps $1.5M); bridge grants still about $6.9M; from Year 4 about $0.31M a year short ($2.2M over Years 4–10); about $1.36 of public money for every private dollar in the campaign, unchanged. National package programme cost about $78.3M a year with the fee reset at the pilot centres only, leverage 1.14 (cash 1.03). Break points unchanged.
Why. Recomputed with the Buying Circles' costs in part A and the fee reset at the pilot centres only.
Before. T3.2's funding-model notes still described how the ledger was derived from the old prep co-op lines (the earlier combined Taste Council and equipment line, the co-op manager, the co-op site, the split revenue lines).
After. Those derivation notes are deleted; the notes describe the ledger as it now stands.
Why. Uncle James (28 Oct 2026): delete stale notes.
Before. The Lighthouse and the pilot centres were to be centres NEA manages itself, so that one landlord answered for every mechanism; Town Councils deferred (§1.3.1), and the fee reset was an ask of NEA alone, booked as income the State gives up at NEA centres.
After. Any of the 123 government hawker centres can host a pilot centre: 27 NEA-managed (26 open), 80 Town Council-managed and 16 run by socially-conscious enterprise operators (NEA list, 17 Aug 2026). NEA regulates every tenancy and stays the tenancy landlord at the 107 NEA and Town Council centres, so rents, bands, the ballot and tenancies work the same there. A Town Council joins as a partner: it manages the common property and collects the conservancy and table-cleaning fees, so the fee reset at its centre needs its agreement and the income given up is its own. An operator also sets tenancy terms within NEA's framework, so the bands, the ballot, its fees and the reset need its agreement, and any share of fee income it gives up is its own, not the State's. Mixing managers tests the plan under each. Privately owned coffeeshops and food courts are left out for now; one may join only on the board's recommendation and without the Hawker Council's veto (§3.8, §6.2). The fifth place the State is asked for money is now the fee income NEA or a Town Council gives up at the pilot centres. No budget figure or stall-count assumption changes.
Why. Uncle James (28 Oct 2026, 09:42 and 09:53 SGT): "I won't stick to rules that don't comprehensively represent hawker centres; expand the definition to include Town Council-managed ones"; then include the 16 operator-run centres, which are government centres, and exclude only privately owned coffeeshops and food courts.
Before. §6.3 described the Lighthouse as a mid-size, mixed-cuisine centre with a substantial 55+ cohort and named no centre.
After. Seven selection tests: a government centre; footfall; heritage density (MICHELIN Guide or Hawker Heritage Award stalls, stalls of 30+ years or multi-generation); between them Chinese, Malay and Indian cuisine, with halal stalls; a mix of NEA and Town Council management; no major works in the pilot window; the association's support and the §6.2 vote. Geylang Serai Market and Food Centre is included by default, subject to its hawkers' vote; if they decline, the plan seeks another centre strong in Malay cuisine. Examples to consider, each subject to its own vote: Chinatown Complex (a possible first centre), Tekka Centre, Old Airport Road (Blk 51) and Maxwell Food Centre. The paper notes that Chinatown Complex and Old Airport Road are bigger than the 120-stall centre the ledgers cost. Twelve new sources.
Why. Uncle James (28 Oct 2026): "Describe the criteria while naming possible locations as examples to be considered"; "consider the importance of Malay cuisine, so we are racially balanced and inclusive; by default at least include Geylang Serai Market and Food Centre."
Before. One Lighthouse Centre opened in Year 1 and two more pilot centres a year later, Geylang Serai included by default and Chinatown Complex, Tekka, Old Airport Road and Maxwell named as examples. §6.3 was titled "The Lighthouse and the Pilot Centres" and §7.3 "Year 1 — The Lighthouse Launches".
After. Two Lighthouses open side by side in Year 1, Chinatown Complex and Geylang Serai Market and Food Centre, for racial balance and inclusion (Chinese and Malay cooking, halal stalls from the start); Tekka Centre is the third pilot centre and opens in Year 2. Old Airport Road (Blk 51) and Maxwell are examples of a fourth and a fifth, stated add-ons. Every centre is subject to its own hawkers' vote; if one declines, the plan seeks another centre with a similar cuisine strength. Year 0 trust deposits, listening and co-design run at both Lighthouses, and Tekka's in the six months before it opens, in Year 1. The gate is unchanged: no further than five centres until the evaluator certifies the criteria across all the pilot centres, its month 43 counted from when the Lighthouses open. One term throughout: "the two Lighthouses", "the Lighthouses", "a Lighthouse"; Part VI is "Year Zero & the Lighthouse Centres", §6.3 "The Lighthouses and the Pilot Centres" and §7.3 "Year 1 — The Lighthouses Launch", with site redirects from the old routes.
Why. Uncle James (28 Oct 2026): two Lighthouse centres in Year 1, Chinatown Complex and Geylang Serai, side by side for racial balance and inclusion; Tekka the third centre, in Year 2; Old Airport Road and Maxwell as add-ons; rename consistently.
Before. Each later pilot centre was assumed to be a mid-size centre of about 120 stalls and about 70 stallholders aged 55+, at the Lighthouse's ledger rates one year behind: 18 Master Fellowships at the Lighthouse (4, 6 and 8) and a flat 10 at each later centre (4, then 6), 20 apprentices, about 13 dishes and 5 landings; each further centre added $2,580,700.
After. Each centre is costed from its own shape. Heritage-worthy stalls, a proxy counted on 28 Oct 2026 (Tier 1: MICHELIN Bib, Selected or Plate stalls, current or former stars, Hawker Heritage Award winners; Tier 2: stalls documented as running 30+ years, or second generation or later; each counted once), padded for guide coverage into a planning count: Chinatown Complex 226 cooked-food stalls, 18 counted, x1.5, 27; Geylang Serai 63, 7, x1.2, 8; Tekka 119, 6, x1.5, 9; Old Airport Road 168, 11, x1.5, 16; Maxwell 103, 13, x1.0, 13. One Master Fellowship ($72,000) per stall in the planning count over the centre's three years, a ceiling, unspent money returning to the endowment, phased 30/35/35% (Year 1 centres) or 40/60% (Year 2 centres): 27 (8, 9, 10), 8 (2, 3, 3) and 9 (4, 5). Two apprentices a Fellowship ($18,000 a year), one documented dish a Fellowship ($18,000), landings for about 30% of apprentices, a second-station grant ($3,000) and $15,000 of first-loss reserve and escrow a Fellowship (the former $570,000 for 38). By stall count: the critical-illness pool (70 in 120 stallholders, $750: 132, 37 and 69), the Corps bench (about 10 cooks a 100 stalls, two-thirds in the first year, $40,000 a bench plus $12,500 a member: 15 then 23, 4 then 6, 8 then 12), circle leads (four a centre at least, one for about every 30 stalls: 8, 4 and 4) and centre staff (1.5 FTE for every 120 stalls, rounded down to the half, at least 1.5; Chinatown Complex adds one beyond the core team's Lighthouse roles). Revenue $10,000, $14,500 and $25,000 a Fellowship by the centre's age. Each Lighthouse buys a $30,000 capture kit and pilot-documents three hawkers in Year 0. A new T1.4.6 sets out the proxy, its padding and its sources (three new).
Why. Uncle James (28 Oct 2026): "consider the shape of each hawker centre; how many are likely to be heritage/conservation worthy? Use proxy data, e.g. popularity ... as proxy inputs ... with some padding as budget-per-centre logic."
Before. $12,707,110 over Years 0–3 (Y0 $526,900, Y1 $2,275,170, Y2 $4,335,570, Y3 $5,569,470); operating $4,448,620, programme $8,258,490 (65.0% reaching hawkers); burn about $92,700 a month; 38 Fellowships, 76 apprenticeships, 23 landings; bridge grants for gaps of about $6.9M; from Year 4 about $5.65M a year, $0.31M a year short ($2.2M over Years 4–10, $12.1M over Years 4–20); corpus keeps its real value in 50% of futures; $1.36 of public money per private dollar; downside, the Lighthouse alone, $2,716,170 in Year 3 ($7,288,310 over four years); first-loss budget $570,000.
After. $14,628,275 (Y0 $824,725, Y1 $3,479,950, Y2 $4,895,500, Y3 $5,428,100); operating $4,910,840, programme $9,717,435 (66.4%); burn about $102,300 a month; 44 Fellowships, 88 apprenticeships, 26 landings, 238 stallholders 55+ covered; gaps of about $8.5M ($2,459,950, $3,045,500, $3,028,100), Year 0 foundation grants $631,725; from Year 4 about $5.43M a year in Year 3 prices ($5.51M in Year 4), $0.16M a year short ($1.1M over Years 4–10, $9.4M over Years 4–20); corpus keeps its real value in 46% of futures; $1.27 of public money per private dollar. Add-ons: Old Airport Road $3,888,980, Maxwell $3,072,400 (four centres $18,517,255, five $21,589,655); a further centre not yet named (100 stalls, 10 heritage-worthy) $2,587,250. Downside: both Lighthouses keep running and Tekka defers, $4,135,775 in Year 3 ($12,138,450 over four years), no later year short. First-loss budget $660,000 ($560,000 after escrow set-up). The Monte Carlo now draws Tekka's veto alone (80%); a Lighthouse that declines in Year 0 is replaced.
Why. Recomputed from the per-centre ledgers (funding_model.py, in parity with funding.js).
Before. The fee reset at the pilot centres assumed about 120 stalls a centre, about $0.86M a year at three ($288,000 a centre) and $1.44M at five; the national package about $78.3M a year (cash leverage 1.03).
After. Counted at each centre's actual cooked-food stalls, $2,400 a stall a year: Chinatown Complex $542,400, Geylang Serai $151,200, Tekka $285,600, $979,200 at the three (408 stalls) and $1,629,600 at five (679 stalls, with Old Airport Road and Maxwell); the national package about $78.4M a year (leverage 1.14, cash 1.02).
Why. Uncle James (28 Oct 2026): lines that follow the stall count use each centre's actual count.
Before. No comment on how the flat allocation compared with each centre's heritage.
After. "The flat allocation was about right in total – 38 Master Fellowships, 18 at the one Lighthouse and 10 at each later centre, against the 44 the heritage count asks for – but in the wrong places: nine short at Chinatown Complex, two too many at Geylang Serai and one at Tekka. Yet money is not the real limit. Willing masters are: NEA's Hawkers Succession Scheme completed six handovers between January 2022 and September 2025 (MSE, 22 Sep 2025). So the Fellowship line is a ceiling, not a target, and what a centre does not spend on it returns to the endowment."
Why. Uncle James (28 Oct 2026): say that the flat allocation was about right in total but in the wrong places, and that willing masters, not money, are the real limit.
Before. §3.6 Endangered Dish Registry, a ranked register of dishes kept with NHB, culinary historians and communities; no stall-level tier and no yearly count.
After. A light first tier, the Heritage Watchlist, national from Year 1 at all 123 government hawker centres. A stall qualifies on any one test (20 years or more at the stall; a second generation or later; MICHELIN, the Hawker Heritage Award or an NHB or NEA heritage board; or a nomination by its hawkers' association or by 20 regulars, an assumption), confirmed by one short visit; private coffeeshops and food courts by the existing exception route. A listed stall gets a public listing and a counter decal ("On the Heritage Watchlist"; recognition, not a quality mark, so it does not compete with the Hawker Heritage Award or the Heritage Plate), a story-and-dish capture for the open repository and first call on Relief-Corps days. A yearly check-in with consent asks four questions (age band, named successor, tenancy end, health); only aggregates are published, in line with the PDPA. Two or more signals (owner 65 or over, no successor, tenancy ending within three years, among the last stalls cooking the dish) flag a stall at risk: it goes to the front of the queue for a Master Fellowship and full documentation, and its dish feeds the Registry. Each year the count of stalls watchlisted and at risk, by centre, cuisine and community, goes to NEA and MSE, NHB and MCCY, and the Singapore Tourism Board, from the end of Year 1; §7.7 adds a zero-cost row asking them to use it. Assumed 300, 600 and 800 stalls by the end of Years 1–3 (12–15% of the 6,000+ cooked-food stalls), 15–25% of them at risk (about 120–200 by Year 3). Cost in the operating budget, because its stipends pay visitors, not hawkers: $99,000, $107,000 and $114,000 in Years 1–3 before contingency (a 1 FTE coordinator; 1,700 visits at a $30 stipend and $10 travel; $10 decals; the listing on the open price board, the increment only), $352,000 with contingency.
Why. Uncle James (28 Oct 2026): "a lite version to pre-qualify a stall as heritage worthy so we can encapsulate more stalls … you're in my watchlist", so the plan can move quicker to the stalls truly at risk and give Parliament and its partners (NEA, the Singapore Tourism Board and others) an at-risk count.
Before. §3.6 Endangered Dish Registry.
After. §3.6 The Heritage Watchlist and the Endangered Dish Registry, not renumbered; the Watchlist first, as the stall-level front door, and the Registry second, as the dish-level ranking. Its old web route redirects.
Why. Uncle James (28 Oct 2026): the Watchlist is the first tier of §3.6.
Before. Fig. ES.1 mapped sixteen needs; one row was empty (repairs and exhausts, left to landlords). ES.2's sixth cell was left empty.
After. Fig. ES.1 maps seventeen: "Knowing who is at risk" under Handing over, keeping the dish, with nothing found today (MCCY, 7 Oct 2026, keeps no list; NHB's inventory of intangible cultural heritage names practices, not the stalls at risk), and the plan filling the gap (§3.6, paid from gifts and the endowment). ES.4: two rows are empty; the plan fills one, knowing who is at risk, and leaves the other to the landlords. ES.2: the Watchlist and the Registry would fill the sixth cell.
Why. Uncle James (28 Oct 2026): show the new need in the map, honestly marked.
Before. $14,628,275 over Years 0–3 (Y0 $824,725, Y1 $3,479,950, Y2 $4,895,500, Y3 $5,428,100); operating $4,910,840, programme $9,717,435 (66.4%); burn about $102,300 a month; gaps of about $8.5M; listening fees 150 at every centre; Old Airport Road $3,888,980, Maxwell $3,072,400; the two Lighthouses alone $12,138,450 ($4,135,775 in Year 3); $1.27 of public money per private dollar.
After. $14,990,340 (Y0 $834,955, Y1 $3,588,685, Y2 $5,013,200, Y3 $5,553,500); operating $5,262,840, programme $9,727,500 (64.9%); burn about $109,600 a month; gaps of about $8.9M ($2,568,685, $3,163,200, $3,153,500), Year 0 foundation grants $641,955. Listening participant fees follow the stall count, about 1.25 a cooked-food stall (150 for the 120-stall centre the ledgers were first drawn for): 283 at Chinatown Complex, 79 at Geylang Serai (was 150 for 63 stalls), 149 at Tekka. Old Airport Road $3,898,880, Maxwell $3,068,935 (four centres $18,889,220, five $21,958,155); a further centre $2,583,125. The two Lighthouses alone $12,500,680 ($4,261,175 in Year 3). $1.26 of public money per private dollar.
Why. Uncle James (28 Oct 2026): the Watchlist's cost, and the open item that Geylang Serai's listening carried 150 participant fees for 63 stalls.
Before. From Year 4 each centre kept its Year 3 budget, indexed, so Chinatown Complex funded ten new Fellowships a year beyond its 27 heritage-worthy stalls; about $5.51M a year at three centres, $0.16M a year short ($1.1M over Years 4–10); every continuing gift spent; the corpus kept its real value in 46% of futures and ended Year 20 at 86% of target; retirement-match shortfall over Years 4–10 $15.7M base, $33.6M stretch with HDB ($38.6M without).
After. From Year 4 the steady state follows the heritage logic: once a centre's planning count has been offered, new Fellowships continue at a maintenance rate fed by the Watchlist's at-risk flow, an assumption of about 10% of the planning count a year, at least one (3 at Chinatown Complex, 1 at Geylang Serai, 1 at Tekka; 2 at Old Airport Road, 1 at Maxwell), with their apprentices, dishes, grants, reserve and landings; stall-driven lines and the national core stay at their Year 3 level. About $3.50M a year at three centres in Year 3 prices ($1.02M, $0.51M, $0.62M; add-ons $0.89M and $0.58M), against about $5.31M from the draw, the operating match and continuing gifts: no year short, a little under half of continuing gifts spent, the corpus keeps its real value in 89% of futures and ends Year 20 at 160% of target. At 20% a year it ends at 132%, at 30% at 103%, and at 40% three centres run $0.8M short over Years 4–10. The two Lighthouses alone $2.88M a year from Year 4. Retirement-match shortfall over Years 4–10 with half of surplus gifts set aside: $8.8M base (short from Year 7), $26.8M stretch with HDB ($31.7M without); the corpus then ends Year 20 at 118% of target.
Why. Open item from the last pass: the Year 3 envelope kept funding Fellowships beyond each centre's heritage-worthy stalls; Uncle James (28 Oct 2026) asked for the steady state to follow the heritage logic, fed by the Watchlist.
Before. The Executive Summary's cover said "in seven pages", but the 1.1 additions run to eight.
After. Uncle James accepted eight pages (28 Oct 2026). The cover now says "in eight pages"; ES.7's four figure tiles sit on one row.
Why. The cover states the length the summary actually has.
1.0 27 October 2026 1 change · shared version
First edition. Hawker Heritage is published as three papers that share one version: the Executive Summary (why and what), the Main Paper (what and how) and the Technical Paper (how it holds and how to vary it). This entry is the baseline against which every later change is recorded.
Before 1.2 the plan was three papers, the shared ancestors of both stalls; this entry is in both logs.
Approved by Uncle James · tag v1.0 · release commit 4ada8d6
Before. —
After. The full text of all three papers as released on 27 October 2026.
Why. First release to readers.
Stall 03
Full House
A hawker-centre game. Run a centre for twenty years and live with the trade-offs. In development. Its change log is kept with the game, and opens when the game does.