2026-09-28 (4) — Lint: the nineteenth housekeeping pass
What this run did
This was the fourth and last cycle of the scheduled run that began at 00:44 UTC. A "cycle" is one unit of work that is checked, recorded, and merged on its own. The first three cycles were reviews and merged as pull requests #110, #111, and #112. Five runs had gone by since the last lint pass, so the mode selector (tools/next_mode.py) forced this one.
A lint pass is housekeeping. Scripts check the whole dictionary and its notes. Then I read their output and the project's own pages for problems a script cannot judge. This pass found nothing that needed repair.
What was checked, and what it found
- Word caps and links. Every page is within its word limit, and every relative link resolves.
- Cross-references. When entry A lists entry B as a synonym or a word to compare, B should list A back.
tools/crossref.pyadds those mirror links, but it guesses the wrong sense of B about one time in five, so each new one is checked by hand. This pass it wrote none. The 1,794 targets it reports as missing are words named in cross-references that have no entry yet; they are already in the queue of words to write. - Vocabulary. A second script checks every definition against the defining vocabulary, the list of plain words definitions must use. It found no violations.
- Stale claims. A claim file is how a run reserves headwords so that two runs do not write the same word. There were no stale ones.
- Reviewer precision. Each finding from the two reviewer models is either applied or rejected, and precision is the share applied. When a reviewer falls below 30 percent on one kind of finding over twenty or more decisions, it stops being asked about that kind. No new kind crossed the line; the five switched off earlier stay off. One is about to cross: the first reviewer's findings on inflections (the forms of a word, such as ran for run). It has had 1 of 19 accepted, so one more rejection will switch it off at the next lint pass. The figures are in
wiki/notes/reviewer-precision.md. - Index, open questions, and the curator list. The wiki index lists every page. The open questions still carry their working assumptions.
reviews/needs_curator.txt, the list of items for you, has no duplicate lines.
Two recurring objections recorded as settled
Two reviewer objections kept coming back in today's three review cycles, and I rejected them each time. I have added both to the "fences" in NEXT.md, the list of settled points that later runs should not argue over again:
- The second reviewer objects that the one-sentence summary at the top of an entry (the core idea) is written as a sentence. The style guide requires a sentence there.
- The first reviewer objects to the grammar marker "not used in continuous tenses" on need and own. The marker describes normal use, and each entry's learner-error note already says "not normally".
The second reviewer's objection to the quoted headword at the start of translators' notes also continues, 18 times today. It is already a fence.
Spend
US$0. A lint pass makes no paid calls. The day's total is US$1.69 of the US$15 cap.
What did not work
Nothing failed.
This run as a whole
- Four cycles, all merged.
- Three were reviews, covering 45 verbs written late on 2026-09-26, from need to reach.
- One was a lint pass.
- About 20 senses, subsenses, and phrases were added by hand (for example read meaning interpret, reach meaning extend, play dead, and I'll pass).
- 98 reviewer issues were decided: 43 applied and 55 rejected.
- 315 entries still have only one review round.
What is next
The next review should take the last four late-cycle verbs (prove, provide, pull, punish) and then the nouns and adjectives added from the dictionary's own definitions late on 2026-09-26. The next lint pass is due in five runs.
What needs the owner
Nothing new. Open questions 4, 5 and 6 in wiki/open-questions.md are still waiting for you.