Asked ledger — one row per question sent out, and what came back
builds_on: and related:Asked
One row per question actually sent in an intake, harvest, extraction or mining prompt, and what came back. The mirror of notes/harvested.md: that ledger records that a chat was harvested, this one records that a question was answered — or wasn't. sync_cases.py check_asked_ledger checks it both ways.
Why it exists. The return-path audit (2026-09-21, generalised in N-022 §8) found that the project's extraction prompts were not the failure — the return path was. lemma-harvest §3 asked whether the greedy theorem holds for any non-negative-deficit tie-break; the answer came back as the broader class; the filed statement did not move for a month, and nothing noticed, because nothing was watching for an answer. Seven further questions had never landed at all.
empty is a real outcome, not an omission. A question can legitimately have no answer — §5 of a harvest may genuinely be empty. A ledger that cannot distinguish "nothing came back" from "nobody looked" is worthless, which is the same move docs/OUTTAKE.md makes with a blank line: a blank says "nothing of this kind", and that is information.
Outcomes
| outcome | means | checked how |
|---|---|---|
filed:<path> or filed:<path>#<anchor> | the answer landed | the path must exist; if an anchor is given it must appear in the file |
empty | the question was asked and genuinely had no answer | terminal; nothing to resolve |
open | asked, not yet answered or not yet filed | terminal, but reported as stale past 90 days |
The ledger
| prompt · § | chat UUID | question, one line | outcome | date |
|---|---|---|---|---|
lemma-harvest §1 | necklace-greedy | the one-sided lemma — quantity bounded, direction, hypothesis, proof | filed:statements.js#st/greedy-midpoint | 2026-09-20 |
lemma-harvest §2 | necklace-greedy | what must counts satisfy for midpoint exactness — each even, all equal, something else | open | 2026-08-19 |
lemma-harvest §3a | necklace-greedy | does either result require x = 3, or hold for general x | open | 2026-08-19 |
lemma-harvest §3b | necklace-greedy | is the broader non-negative-deficit tie-break class proven or conjectured | filed:statements.js#st/greedy-midpoint | 2026-09-20 |
lemma-harvest §4 | necklace-greedy | anything left informal — hand-waved steps, cases checked by example | open | 2026-08-19 |
formula-harvest §1 | necklace-discrepancy | what exactly is the arrangement w, and what builds it | filed:statements.js#st/greedy-midpoint | 2026-09-13 |
formula-harvest §2 | necklace-discrepancy | the exact definition of D as computed | filed:statements.js#st/necklace-discrepancy-boundary | 2026-09-13 |
formula-harvest §3 | necklace-discrepancy | the exact definition of c(w) | filed:statements.js#st/necklace-discrepancy-boundary | 2026-09-13 |
formula-harvest §4 | necklace-discrepancy | the precise status of D ≤ 4/3 ⟹ c(w) = 1 | filed:disagreements.js#authored/greedy-midpoint-d-conflation | 2026-09-13 |
formula-harvest §5 | necklace-discrepancy | does D / c(w) / the threshold behave differently for k = 2 vs k > 2 | open | 2026-09-13 |
harvest-prompt §5 | 72cc6f92-e583-4b1c-82a5-591c1be06e23 | questions asked that never got answered (genesis chat) | open | 2026-09-13 |
harvest-prompt §5 | c3333105-7461-4fbc-ad90-549cba7566c0 | questions asked that never got answered (Navier–Stokes chat) | open | 2026-09-13 |
harvest-prompt §5 | 90bc0a56-3620-4e8c-8b62-87d605ca9b2e | questions asked that never got answered (twin-prime chat) | open | 2026-09-14 |
harvest-prompt §6 | 72cc6f92-e583-4b1c-82a5-591c1be06e23 | threads that stopped, and why | open | 2026-09-13 |
harvest-prompt §7 | 72cc6f92-e583-4b1c-82a5-591c1be06e23 | the assistant's own read | empty | 2026-09-13 |
mining-5 | five-deep-chats | orphaned arguments, post-hoc corrections, load-bearing speculation across five chats | filed:notes/rediscovery-inventory-e8.md | 2026-09-21 |
mining-5 | five-deep-chats | the selection-effect cross-chat evidence ledger | filed:notes/selection-effect-probe.md | 2026-09-16 |
mining-6 §14 | necklace-greedy | the y-scaling result and its Stage-3 exact-checkpoint mechanism | open | 2026-09-20 |
mining-6 §1 | necklace-greedy | the sphere/squaring Job 1 / Job 2 split with the ℓ¹-ball counterexample | open | 2026-09-20 |
The mining-5 row is resolved — 2026-09-21 — and it is worth recording how. It was seeded deliberately as a filed: row pointing at a path that did not exist, so the check's first run would have something real to find. It did. But the answer was not missing: it was misfiled. The inventory had been written on 2026-09-16 into Claude's memory store instead of the repo, where no repo check can see it. The ledger reported it missing from the repo and was right about exactly that, while the content existed elsewhere the whole time.
The check was not wrong; "filed" was ambiguous. It can mean written somewhere durable or written where the checks read, and only the second is what this ledger asserts. The row's date is now the date it reached the repo (2026-09-21), not the date it was written (2026-09-16); the note's own frontmatter carries both. Generalised into OPS.md, an answer can be filed to the wrong store.
71 source lines, no addenda.