2026-10-16, cycle 2 of the 12:44 UTC run: the seventy-fifth lint pass
What this cycle did
This was a lint cycle: a periodic health check of the whole dictionary and of the project's knowledge base (the wiki), rather than new or revised entries. The scheduler (tools/next_mode.py) forced it, because five cycles had passed since the last lint pass. A lint pass costs nothing: it makes no calls to the paid outside models.
It has two halves. The mechanical half runs the project's checking tools over everything. The judgmental half reads the wiki's catalog, log, and open questions to make sure they agree with each other and with the files.
The mechanical checks
- Size caps (
tools/check_caps.py): every capped file is within its limit. - Links (
tools/check_links.py): every link between the project's pages works. - Cross-references (
tools/crossref.py --all --apply): when entry A points to entry B as a related word, B should point back to A. The tool found one missing back-link and added it: near (the preposition) names nearly in its word-family box, so nearly now names near (the preposition) in return. That is right; the word-family box belongs to the whole entry, not to one sense, so there was no risk of the link landing on the wrong sense, which has happened with other kinds of back-link before. The tool also reported, as it has for several passes, that lonely calls alone a word to compare while alone calls lonely a synonym. That stays: the two entries are talking about different senses of alone, and both are correct. - Vocabulary (
tools/lint_vocab.py --all --queue): definitions are meant to use a fixed list of plain defining words, or words that have their own entry. The check queued no new words. - Claims (
tools/claim.py --prune): two old claim files (records that a cycle had reserved some headwords) from the morning's build cycles were removed, since those words are finished.
Reviewer precision
The lint pass measures, for each reviewer and each kind of issue, how often its objections turned out to be right (the share of its issues that I or earlier sessions applied rather than rejected). A kind of issue that falls below 30 percent over at least twenty decisions is switched off for that reviewer, because it costs more reading than it saves. Nothing new crossed that line. reviewer-a is right about 72 percent of the time overall, reviewer-b about 53 percent. reviewer-b's spelling-and-format objections are below the line (28 percent) but stay switched on by the owner's ruling of 2026-10-05. Its objections about idioms and phrases, at 31 percent, are the closest to the line. In the five cycles since the last lint pass, 6 of reviewer-b's 16 objections were applied. The figures are in wiki/notes/reviewer-precision.md.
The judgmental checks
- The wiki's catalog (
wiki/index.md) lists every page, and every page it lists exists. - The log (
wiki/log.md) has an entry for every cycle since the last lint pass. - The two open questions for the owner in
wiki/open-questions.md(continent names in definitions; whether the former and the following standing alone are nouns or pronouns) are still open, with their working assumptions in force. - The list of items for the owner (
reviews/needs_curator.txt) has no duplicate lines. - I checked the words that
NEXT.md(the hand-over note between sessions) says are linked but have no entry yet, such as then (adverb), usually, and nowadays; they are still missing and still queued, so those lines are current. The two new missing targets from this run's review cycle, nanny and bolt, are in the queue.
Spend
None. The day stands at about US$4.67 of the US$15 cap.
What is next
The next lint pass is due in about five cycles. The scheduler will choose the next unit; the oldest entries still read only once by the reviewers start with official, oil, and operation.
Needs the owner
Nothing new.