lab/runs/*.md with sealed:, no outcome:, no blocked_on: and no running: — a run with compute in flight is Running, not here
any record with a non-empty blocked_on: or state: draft — the sentence in it is the display text
running: <machine or chat>, <since> — set when compute starts, removed at scoring
probes/*.md with state: open, oldest first
decision: — a question only the owner can answer, removed when answered
A document R is stale on record X when R cites X by id and X’s state:, outcome:, resolved: or retired: changed in a commit later than R’s own last commit. Both sides are git commit instants, parsed before they are compared; the changed field is found by reading X’s frontmatter at each commit, not by whether the file was touched. A record whose condition never changed returns its creation, and then the line says created, never scored. Where the two instants share a local day the time is shown too.
bench.md:280same side twice; 2, 4 and 4 of each). Those three kinds are the table run 097 records.
docs/BANDS.md:752026-10-07 — filed. GUARD.WAIT at 6, not the mockup's 18: on run 105 the guard waited 30 times against a thresholddocs/DATABASE-AUDIT.md:29rows are run 100's, pre-registered for an open run — the rule working as written.
docs/DATABASE-AUDIT.md:35rowsand is loaded by hand; run 101'srun101.jsonandgates101.jsonhave none. Twenty-two scored
docs/ENGINEERING-LEG.md:40It would have caught the L=28 NaN before the run.** Run 023's L=28 leg returned a NaN because the transfer operator had no eigenvalue on that grid — a substrate refusal, not a verdict, and the tool misreported it as `FAI
docs/PROGRAMME-AUDIT.md:51PG-02 objects | seam, free coset, kind (two senses), flippable, independently flippable, exact part, rest/remainder, frame, family, orbit, trap, W_n, dust, layer, fibre of the level map, twin, basin, stratum | run 101's
docs/PROGRAMME-AUDIT.md:648line when a queued case's run has started drawing or will report within a day — *"CASE-31 is queued and
docs/PROGRAMME-AUDIT.md:649run 102 reports tomorrow; press before then."* It sends nothing when the queue is empty. The task is created
docs/PROGRAMME-AUDIT.md:797control (i)–(v) | 01, 03, run 100:13| used, never defined as a term | a numbered pre-condition whose failure withdraws or de-scores |OPS.md:241statistical control;:490positive control |
docs/REPO-AUDIT.md:59runs 095 and 100, and is enforced by hand ("files over 5 MB"), so run 100's five 48 MB files sit
docs/REPO-AUDIT.md:315(c22840a), so run 101'sscript:line could carrymain()again. 36 of 36 node checks,python3 -m tests
docs/RUNG-AUDIT.md:65§8.2 mechanism instances — Tsirelson 2.8284271, sphere-packing d=8/24, bootstrap, Motzkin/SDP, run 025 | R | → new notes/N-0NN-bound-mechanism-instances.md. |docs/compute-environment.md:63M5** | MacBook Pro, Apple M5, arm64 | 10 (4 performance, 6 efficiency) | 16 GiB | 5cf50dc6a4ede1d1abbf6082bbacfbdc | H7 (2187): 30.4 samples/s at 9 threads (run 100); H9 (19,683): 0.178/s at 2 threads, 5.99 GB peak (rudocs/compute-environment.md:101BernoullisBook's slowing from piece to piece in run 102 (57,198 → 61,792 s) is the same instructions at more
docs/design/2026-09-21-make_site.md:111Keying on the state alone mislabels run 023, which isopenwith ablocked_on.
docs/design/2026-10-01-make_site.previews/K-programme-big.html:179<div class="line"><span class="lid"><span class="id">S1</span></span><span style="flex:1;max-width:34rem"><div class="body-tight"><b>Block independence.</b> The one-block map: a block's crossing is independent of it
docs/design/2026-10-01-make_site.previews/K-programme.html:179<div class="line"><span class="lid"><span class="id">S1</span></span><span style="flex:1;max-width:34rem"><div class="body-tight"><b>Block independence.</b> The one-block map: a block's crossing is independent of it
docs/design/2026-10-07-resonance-console.md:17programmes, one draw in flight on the M5 (run 105, piece 1 of 8) and one sealed and not started on
docs/design/2026-10-07-resonance-console.md:18BernoullisBook (run 102). Every value on every screen was read off disk that morning and is printed
docs/design/2026-10-07-resonance-console.previews/A-L1-alarm.html:127<div class="alarmline"><svg viewBox="0 0 16 16" width="13" height="13" style="color:#96241f;flex:none;margin-top:4px"><path d="M3.2 3.2 12.8 12.8 M12.8 3.2 3.2 12.8" stroke="currentColor" stroke-width="2.6" stroke-lineca
docs/design/2026-10-07-resonance-console.previews/A-L1-large.html:127<div class="alarmline quiet"><svg viewBox="0 0 16 16" width="13" height="13" style="color:#1d6440;flex:none;display:block" aria-hidden="true"><circle cx="8" cy="8" r="6.4" fill="currentColor"/></svg><div class="body"><sp
docs/design/2026-10-07-resonance-console.previews/A-L1-normal.html:127<div class="alarmline quiet"><svg viewBox="0 0 16 16" width="13" height="13" style="color:#1d6440;flex:none;display:block" aria-hidden="true"><circle cx="8" cy="8" r="6.4" fill="currentColor"/></svg><div class="body"><sp
docs/design/2026-10-07-resonance-console.previews/B-L3-m5.html:127<div class="alarmline"><svg viewBox="0 0 16 16" width="13" height="13" style="color:#96241f;flex:none;margin-top:4px"><path d="M3.2 3.2 12.8 12.8 M12.8 3.2 3.2 12.8" stroke="currentColor" stroke-width="2.6" stroke-lineca
docs/design/2026-10-19-transfer-operator-svd.md:59Built by a tool chat on the M5 across two sessions: paused on 2026-10-19 (22:00) for run 107's compute after 44 jobs, whose outputs were lost because the script wrote only at the end (the brief never said "write per job"
docs/design/2026-10-21-theme-layer.md:43symmetry | present / broken / hidden | the block (PG-01's four absences) | the transfer operator's T = M Tᵀ M check; the Laplacian's irreps (run 108) |
docs/design/2026-10-22-symmetry-ladder.md:55Every census pattern is D4-symmetric.**census068.pylists symmetric patterns only; the 1/4/8 orbit sizes the codifier remembered are the cell orbits of the 15-cell wedge, not pattern orbits;census068.jsonhas no
docs/design/2026-10-24-db-catch-up.md:68the loader ingests what a RESULTS file states and computes nothing. **From run 135 on, both families should
docs/explainer-carpets-pictures.html:105<section class="rec" id="p1"><div class="app"><b>PLATE 1</b><span>the objects</span><span>T and X</span></div><div class="body"><h2>Two patterns carry the whole programme</h2><p>Each pattern is a 9 × 9 grid with 17 cells
docs/rung-vocabulary-v3.54.md:1445Whether this is the general mechanism — whether the same argument explains why Donaldson invariants are not traces — is exactly CASE-14's open question and is not settled by it. One of the three instances now has a r
docs/rung-vocabulary-v3.54.md:1532C11(a)** | CASE-11 resolves: a Γ-product indexed by class-group characters is not Poisson summation on a lattice |
docs/v5-plan.md:237PN-009 | CASE-31 | the sign of the lean follows minimum-wall size C against shortest-path length l, not the wall count N (found: the diamond (b,s) grid, run 110; the 63 generators' C(1) against l = 8); the second-crossin
docs/v5-plan.md:237PN-009 | CASE-31 | the sign of the lean follows minimum-wall size C against shortest-path length l, not the wall count N (found: the diamond (b,s) grid, run 110; the 63 generators' C(1) against l = 8); the second-crossin
docs/walk/2026-10-21-owner-thread-spectra-classification.md:10236. The blind chat's second answer (19:42). σ on the full n × n grid at p = 0.5 is 2.83e−4, 1.15e−4 and 8.05e−5 for n = 3, 5, 7, with a forecast of about 7e−5 at n = 9. σ(p) peaks near 0.23, not at the fixed point; w
docs/walk/2026-10-21-owner-thread-spectra-classification.md:119Where it costs: the census's 63 are D4 orbits worked by representatives (orbit size never a statistic; run 107's wall-entry classes cut across the D4 classes and the relation between the two quotients was never written);
docs/walk/2026-10-23-today-in-plain-words.md:73Applied to our 63 patterns from their shapes alone, the rule predicts an exact count for each. Run 126 counts them.
docs/walk/2026-10-24-arc-period-three.md:60The "0.84" seen at k ≤ 64 is λ minus an exact 1/√k offset from the half-integer window edges, still climbing. Run 129 tests all four limits at k = 72–128.
docs/walk/2026-10-24-the-comb-in-plain-words.md:34What run 130 asks.** It fits the four lowest teeth all at once, on fresh random draws, on a bigger carpet (L = 2187). Two things must hold:
docs/walk/2026-10-24-the-comb-in-plain-words.md:54Run 129, the four constants at the period-three crisis, is computing on the M5 and will finish this afternoon. Runs 130 and 131 queue behind it, because the M5 has 16 GB.
docs/walk/2026-10-24-today-in-plain-words.md:41What the block operator can't see (formation, then run 132).** Two different cell flips can give exactly the same operator. A reader found the rule. Strip the block's graph to the links that matter: drop dead ends, merge
docs/walk/WALK-README.md:49FLOW.md | D3; the lead leaves the frontmatter (body section "Not part of the seal", lab prompts say "do not read past it"); precision of a prediction scales with the stage (the devil's advocate answered); the intake stprimers.md:115the corrected one. **Run 029's own expectation of 73 levels at R = 50 came from LMFDB directly9 citations skipped: completed checklist items. A - [x] line records what happened; this rule asks what a document still says. The skip is line-level and mechanical, and a ticked line stating the wrong outcome is still caught by the value check below.
R states an outcome word beside a run id that is not the run’s outcome — run 9xx — confirmed, confirmed by run 9xx, run 9xx scored confirmed. Only those shapes. No dates involved.
lab/runs/105-h-19683-prediction.md says falsified beside run 100; the record says confirmed — “falsified by run 100”a literature claim flagged from memory or unverified, with an attribution in the same clause; it waits for a pinned citation. The rule, its drops and the bibliography are on Citations.
unticked items in TODO.md that name no record already listed on this page; the age is the commit of the item’s first line (git blame), and what it waits on is the section it is filed under. Ticked items never appear. 11 excluded as duplicates of rows above: TODO.md:113 (027), TODO.md:129 (P-007), TODO.md:602 (023), TODO.md:76 (P-005), TODO.md:426 (028), TODO.md:521 (023, 029), TODO.md:443 (023, 029), TODO.md:213 (P-043), TODO.md:216 (P-043), TODO.md:648 (135), TODO.md:243 (026).
zeta-zeros-rmt as a course.algebra-geometry-dictionary.md into the repo.structural lens edges.information-theory is broader than the statement that forced it.domain-math.js.lab/index.html into build-all.sh or a git hook so it can't drift.scripts/, and move all or none.liouville_1e9.py before scaling past 10⁹.reading/.~/Math_Map vs ~/mapofmath split — one copy of every script, symlinked.status.sh — one command that shows every active run.status.sh proves worth it] A runs section on the lab page.held, not built.make-course-bundle.sh — one merged PDF per course in courses/src/.prerequisites: on every open case and run, then a query.py view.probes/ has no page, so P-nnn in body text is now plain text.lab/make_site.py is in no automation.python manifest.py --write and a commit of MANIFEST.md alone; the workflow fails otherwise.krasner-degree-growth is held on its sequence-independence clause (CASE-07's follow-up): the same element through a second approximating sequence; BernoullisBook, minutes.19c1f7c…03b1360) left open.166 rows. Sorted oldest first inside each group; the bar fills at three weeks and goes to full ink, so a stale item looks stale without a colour. A doc is not a record: it has no display state and no age, and its row shows an em dash where the bar would be. Docs are read only when they carry decision:, running: or blocked_on:.
generated 2026-10-25nothing on this page is maintained by hand