2026-10-18, cycle 4 of the 12:43 UTC run: the seventy-eighth lint pass
What this run did
This was a lint cycle: a check of the whole dictionary and its knowledge base (the wiki/ folder of rules, notes, and decisions) for problems that build up over time. The scheduler (tools/next_mode.py) forces one every five runs; this was the fifth run since the last pass. A lint cycle makes no paid calls and writes no new entries.
It has two halves. The mechanical half runs the project's checking tools over everything. The judgmental half reads the knowledge base for contradictions, gaps, and stale items that no script can see.
The mechanical checks
- Word caps (
tools/check_caps.py): every capped file is within its limit. - Links (
tools/check_links.py): every link inside the knowledge base points to a file that exists. - Vocabulary (
tools/lint_vocab.py --all --queue): no definition anywhere uses a word outside the defining vocabulary that lacks an entry. One queue row's note was updated; nothing new was queued. - Claims (
tools/claim.py --prune): three old claim files (the lists a cycle writes so that two runs never draft the same word) belonged to cycles whose entries have all merged, and were removed. - Cross-references (
tools/crossref.py --all --apply): this tool adds the "back-link" when one entry lists another as a synonym or a family member and the other does not list it back. It proposed six back-links in four entries. Past passes showed that it sometimes attaches a back-link to the wrong sense, so I checked each one. All are right: upward as a synonym of up in its first sense (from a lower place to a higher place); mostly beside usually; okay beside well in in a good, right, or skillful way; and good in the word families of goodbye and well. Goodbye is listed as a relative of good by form, which is what a word family records, so it stays.
One fix by hand
In the previous build cycle I noticed that the entry for first (the number) pointed its "see also" link at second the unit of time, because second the number did not exist yet. It does now, so I changed the link by hand to point at second-num.
Reviewer precision
The tool tools/metrics.py --precision measures, for each reviewer and each kind of issue, how often I agreed with the reviewer. A kind of issue where I agree less than 30 percent of the time over twenty or more decisions is switched off for that reviewer. No new switch-offs were needed. reviewer-a stands at 71 percent overall, reviewer-b at 53 percent.
reviewer-b's phrase issues stand at 31 percent (19 of 61), one rejection away from the line. In the three cycles since the last pass reviewer-b had 2 of 22 issues applied; twenty of the rejections were one claim repeated on the new greeting entries (that say hello is a noun use), which the previous cycle recorded in wiki/notes/reviewer-noise.md. The figures are in wiki/notes/reviewer-precision.md.
The judgmental checks
- The wiki index (
wiki/index.md) lists every page, and every page it lists exists. - The log (
wiki/log.md) has an entry for every cycle, with no gaps. - The two open questions for the owner (continent names in definitions; whether the former and the latter are nouns or pronouns) are still open, with working assumptions in place.
- The curator file (
reviews/needs_curator.txt) has no duplicate lines.
Spend
None. The day stands at about US$5.80 of the US$15 cap.
What is next
The next lint pass is due in about five cycles. The build queue continues with the interjections oh and um.
What needs the owner
Nothing new.