paper-claims widget and blog post: 45 printed numbers from three format papers, 34 fit, 8 do not (Closes #1689)

[gHashTag]

Closes #1689

What

• Widget paper-claims (specs/widgets/paper-claims.t27, category compiler): a table of the 45 numbers printed in the GoldenFloat (arXiv:2606.05017v5), Golden Ruler (arXiv:2606.09686v3) and TNF papers, each with its recomputed value and verdict, plus a flow diagram of how a number is checked. Totals: 34 fit, 8 do not, 3 not checkable. Every row is one row of the t27_paper_claims MCP answer (t27_paper_claims: recompute the numbers printed in the GoldenFloat, Golden Ruler and TNF papers t27-mcp#23), which runs gHashTag/t27 specs/numeric/paper_claims.t27 (numeric: paper_claims.t27 recomputes 45 numbers printed in the GoldenFloat, Golden Ruler and TNF papers; tri papers (Closes #10392) t27#10401). The spec has 3 tests: the cells fill whole rows, every paper adds up, and the totals are the rows.
• public/widgets/paper-claims/tool.js is the established three-line pattern, a single import '../infographic.js'; it holds no words. card.png is 1200x630, rendered from a static SVG with rsvg-convert on the Railway lab.
• Gallery entry (gallery-add, index 244, WIDGET_COUNT 253) and lesson 1 of the GoldenFloat course (LESSON_ALSO), so S15 holds.
• Generated outputs (widget page, gallery, course, course.generated.ts, queenWidgets.generated.ts): produced on the Railway lab by the repo generators, not hand-edited.
• Blog post forty-five-printed-numbers (EN + RU): the per-paper totals table, the table of the 8 misfits, and two figures (verdicts per paper, the check flow) as static SVG strings.

The 8 that do not fit

Paper	Where	Printed	Recomputed
Golden Ruler	Abstract, clusters	12	13
TNF	Table 1 vs Sec. 10, M = N - 1 - Et	8 of 8	2 of 8
TNF	Eq. (1), TNF16 mantissa	9	11
TNF	Cor. 18, kappa(3)/kappa(2)	1.683	1.682
TNF	Table 24, GF-T golden split	9 of 9	1 of 9
TNF	Table 24, "8 of 9 identical"	8 of 9	1 of 9
TNF	Table 24 vs Sec. 10, 1 + Et + M = N	9 of 9	3 of 9
TNF	Table 23, BNF8 exponent bits	3	5

Checked (Railway lab, node 18.20.4)

• widget-pages-from-spec.mjs --check: 433 tests, 1425 asserts pass
• widgets-from-spec.mjs --check: 253 widgets
• course-from-spec.mjs --check: 243 lesson specs compile clean
• widget-logic.mjs --check: OK
• blog-bodies-check.mjs: every published post of 109 opens in the reader
• blog-cover-versions.test.mjs: 7 pass, 0 fail

Not run on the lab: blog-cover-accessibility.test.mjs (needs react; this post adds no cover) and the TypeScript build. CI covers both.

{
  "version": 1,
  "head_sha": "7e0382f06c1e01a36621e6ee9b9e709643aee0ed",
  "summary": "A widget and a blog post show which of 45 numbers printed in the GoldenFloat, Golden Ruler and TNF papers recompute to the printed value in exact integer arithmetic: 34 fit, 8 do not, 3 cannot be checked. The rows come from the t27 spec specs/numeric/paper_claims.t27 through the t27_paper_claims MCP tool.",
  "changes": [
    "apps/website/specs/widgets/paper-claims.t27: new widget spec, 45-row table, per-paper counts, flow, 3 tests",
    "apps/website/public/widgets/paper-claims/: tool.js (import of ../infographic.js), card.png, generated index.html and spec copy",
    "apps/website/specs/widgets/gallery.t27 and apps/website/specs/course/goldenfloat.t27: gallery entry and lesson link",
    "generated: public/widgets/gallery.t27, public/learn/goldenfloat.t27, src/lib/course.generated.ts, src/lib/queenWidgets.generated.ts",
    "apps/website/src/data/blog: bodies/forty-five-printed-numbers.ts (EN + RU), index.ts entry, posts.ts map entry"
  ],
  "tests": [
    {
      "command": "node scripts/widget-pages-from-spec.mjs --check && node scripts/widgets-from-spec.mjs --check && node scripts/course-from-spec.mjs --check && node scripts/widget-logic.mjs --check",
      "result": "433 tests, 1425 asserts; 253 widgets; 243 lesson specs compile clean; widget logic OK",
      "status": "passed",
      "evidence": "Railway lab, /data/pc/trin at the branch head before the blog commits"
    },
    {
      "command": "node scripts/blog-bodies-check.mjs && node scripts/blog-cover-versions.test.mjs",
      "result": "every published post of 109 opens in the reader; 7 pass, 0 fail",
      "status": "passed",
      "evidence": "Railway lab, /data/pc/trin at 8a59b1299"
    }
  ],
  "limitations": [
    "The Lucas identity is checked to n = 45 only; u64 stops there, and the paper claims n = 256.",
    "Three claims are not checkable without their inputs: the FPGA testbench log, the ml_dtypes run, the Table 23 counting method.",
    "blog-cover-accessibility.test.mjs and the TypeScript build were not run on the lab; CI runs them."
  ],
  "tags": ["t27", "widget", "blog", "GoldenFloat"]
}

🤖 Generated with Claude Code

[tri-claw]

Summary by t27-bees

Checked by the t27 compiler

_{t27c parse, parse-complete and typecheck on each changed .t27 file at 7e0382f. Nothing was generated, built or run. A pass means the file is well formed, not that it is right.}

• apps/website/public/learn/goldenfloat.t27: parses and typechecks.
• apps/website/public/widgets/gallery.t27: parses and typechecks.
• apps/website/public/widgets/paper-claims/paper-claims.t27: parses and typechecks.
• apps/website/specs/course/goldenfloat.t27: parses and typechecks.
• apps/website/specs/widgets/gallery.t27: parses and typechecks.
• apps/website/specs/widgets/paper-claims.t27: parses and typechecks.

Read by the model

This PR adds a paper-claims widget and a blog post that verify 45 printed numbers from three papers (GoldenFloat, Golden Ruler, TNF) by recomputing them in exact integer arithmetic. The widget displays a 45-row table with verdicts (34 fit, 8 do not, 3 not checkable), per-paper totals, and a flow diagram; the blog post mirrors this with additional figures. All rows come from the t27_paper_claims MCP tool running specs/numeric/paper_claims.t27.

Worth a look

• apps/website/specs/widgets/paper-claims.t27 – New widget spec with 45 hard-coded rows. If the upstream MCP tool (t27_paper_claims) or the numeric spec (paper_claims.t27) changes its output format or values, this widget will show stale data without a regeneration step.
• apps/website/public/widgets/paper-claims/index.html and paper-claims.t27 – Generated copies of the spec. The PR says they are "produced on the Railway lab by the repo generators, not hand-edited," but they are committed here; if the generation pipeline diverges from the committed files, the deployed widget will not match the source spec.
• apps/website/src/data/blog/bodies/forty-five-printed-numbers.ts – Blog post embeds static SVG strings for figures. If the widget's flow diagram or verdict chart is updated, the blog post SVGs must be regenerated manually or they will drift.
• apps/website/specs/course/goldenfloat.t27 and apps/website/public/learn/goldenfloat.t27 – Both add "paper-claims" to LESSON_ALSO at index 1 (lesson 1). The diff shows the same change in both spec and generated file; if the lesson index shifts, the link will point to the wrong lesson.
• apps/website/src/lib/course.generated.ts and queenWidgets.generated.ts – Generated registry files modified. Since they are committed, any mismatch between the generator and these committed versions will cause CI/type-check failures until regenerated.

---

_{Read: 14 files, +593 -39, the whole diff up to 60000 characters (the rest was not read). Head 7e0382f. Model: openai-compatible/nvidia/nemotron-3-ultra-550b-a55b. Commands: @t27-bees review | summary | help | pause | resume. Free for every repository.}