2026-10-17, cycle 4 of the 00:44 UTC run: the seventy-sixth lint pass
What this cycle did
This was a lint cycle: a check of the whole dictionary and its knowledge base (the wiki) rather than new work on entries. The scheduler (tools/next_mode.py) forces one when five runs have passed without one, and that was the case. A lint pass has two halves: mechanical checks run by tools over every file, and a reading pass in which I look for contradictions, gaps, and stale notes.
The mechanical checks
- Size caps on the rule files, the style guide, and the wiki: all within limits.
- Links between wiki pages and the rule files: all resolve.
- Vocabulary queue: the tool that finds words used in definitions but not yet in the dictionary queued nothing new.
- Claim files (the small files that reserve headwords for a cycle while it drafts): three spent ones were removed.
- Back-links: when entry A names entry B as a synonym, an opposite, or a word to compare, entry B should point back to A. The tool (
tools/crossref.py) added ten such mirror links, most of them because three adverbs drafted earlier in this run (really, sometimes, somewhere) now exist, so links that pointed at them from older entries could be mirrored. The tool always attaches a mirror to the first matching place it can, so I checked each one by hand. Nine were on the right sense. One was not: very compares itself with quite, and the mirror landed on the sense of quite that means unusual (quite a party). I moved it by hand to the sense fairly or very, where the comparison belongs. - One known mismatch remains: lonely calls alone a word to compare, while alone calls lonely a synonym. The seventy-third pass decided this is right, because the two sides name different senses of alone, and it stays.
Reviewer precision
Each lint measures how often each outside reviewer's objections turn out to be right, by kind of objection, over all decisions so far. Any kind of objection from one reviewer that is right less than 30 percent of the time, over twenty or more decisions, is switched off for that reviewer. Nothing new crossed the line. reviewer-a is right about 71 percent of the time overall (7,672 of 10,758 decisions) and reviewer-b about 53 percent (1,326 of 2,499). reviewer-b's spelling-and-format objections (28 percent) stay on by the owner's ruling of 2026-10-05; its phrase objections (31 percent over 59) are the nearest to the line. Eight pairs stay switched off, as before. reviewer-b was very quiet this run: no objections on twenty reviewed entries and two on twelve new ones. The figures are in wiki/notes/reviewer-precision.md.
The reading pass
- The wiki index lists every page, and every page it lists exists.
- The log has an entry for every cycle since the last lint, with no gaps.
- The two open questions for the owner (continent names in definitions; whether the former standing alone is a noun or a pronoun) are still open and still accurate.
- The curator queue (
reviews/needs_curator.txt, the list of items for the owner) has no duplicate lines.
Spend
None. The day stands at about US$2.34 of the US$15 cap.
What is next
This run ends here: it has used its four cycles. The next run's scheduler will choose again; the next lint is due in about five cycles.
What needs the owner
Nothing new.