2026-10-08 (21): lint — the fifty-fifth pass, five back-links, one moved
Cycle 1 of the 20:43 UTC run. Run id 20261008T204353Z-0b154c. This is the run's only cycle: the day's budget is spent.
What this cycle did
A lint pass is unpaid housekeeping: mechanical checks across the dictionary, then a read of the notes for contradictions, gaps, and stale items. The scheduler (tools/next_mode.py, the script that picks each cycle's job) forces one after five cycles without a lint, and this was the fifth: the last cycle of the 12:43 run and all four of the 16:44 run.
No unfinished work from earlier runs was waiting, and the owner's inbox folder was empty.
The mechanical checks came back clean:
- Word caps. Every capped file is within its limit.
- Links. All relative links in 432 notes and reports resolve.
- Defining vocabulary. Definitions must use words from a list of about 2,400 plain words, or words that have their own entry. No definition breaks this rule. One new row was added to the queue of words to write: hips, used in the definition of pants. It is a plural of hip, already queued, so I marked it a duplicate.
- Claims. A claim file records which words a run is drafting, so that two runs never draft the same word. Two claim files belonged to work that has already merged, and they were removed.
Back-links
When entry A names entry B as a synonym, an opposite, a word to compare, or a family member, a tool adds the reverse link to B. This is a "back-link". The tool guesses which sense of B the link belongs on, and it has often guessed wrong, so every lint checks each new back-link by hand.
This pass added five back-links to three entries written today:
- supply sense 1 (an amount available) now lists bank as a synonym, mirroring bank's sense "a supply of something, such as blood, kept for use". Right sense.
- strength sense 1 (the body's physical power) now lists energy and force. Both mirror senses about physical strength. Right sense.
- strength also got power, but the tool put it on sense 1. The power sense it mirrors is about the force of things, as in the power of the waves. That belongs on strength sense 4, "how great or powerful something is, for example a wind". I moved it.
- stuff sense 1 (things in general) now lists shit as a synonym, mirroring the vulgar entry's sense "things of any kind". The link is correct, but a learner reading stuff would see a vulgar word with no warning. I added a short note by hand: "a vulgar word; do not use it in polite speech".
Reviewer precision
Two outside reviewer models check every entry: reviewer-a and reviewer-b. For each kind of issue a reviewer raises, the project tracks its "precision", the share of its issues that I judged right and applied. A kind whose precision stays under 30 percent over twenty or more decisions is switched off for that reviewer.
No new kind fell under the line, and the seven pairs already switched off stay off. Reviewer-a's overall precision is 0.72 (5,511 of 7,703 decisions applied). Reviewer-b's is 0.53 (1,104 of 2,067). Reviewer-b's spelling-and-format issues stand at 0.28, under the line, and stay on because you asked for that.
One trend is worth watching. Since reviewer-b moved to a stronger model on 2026-10-05, 94 of its 198 issues have been applied. But over the 42 entries reviewed since the last lint, it raised only 15 issues, and 3 were applied. That is 0.07 useful issues per entry; at the 2026-10-07 lint it was 0.24. Two builds of twelve entries each drew 2 and 0 issues from it. It still answers every field; it may be finding less to fix, or it may have become too lenient. Under the rule against new machinery on speculation, I recorded the figure in wiki/notes/reviewer-precision.md and changed nothing. If the next lint finds it still under 0.10, the next step is a written note, before any change to its prompt.
The read-through
The page index lists every page; the log has an entry for every cycle today; there are no open questions and no duplicate lines in the curator list (items only a person can settle). 36 one-sense entries still carry a sense label meant only for entries with several senses, as the notes say. I found no contradictions between pages.
Spend
None. The day's spend stays at US$14.82 of the US$15 cap. A run ends early when less than US$0.50 of the day's budget is left, so this run stops after this cycle. The budget resets at midnight UTC, and the next run (about 00:43 UTC on 2026-10-09) can build again.
What is next
The next lint is due in about five cycles. The next run will most likely build nouns from support onward or give a second review to the oldest once-reviewed entries.
For the owner
Nothing new needs a decision. The three disputed pronunciations (several, medicine, outside) still wait in reviews/needs_curator.txt. Reviewer-b's spelling-and-format issues remain at 0.28, under the switch-off line, and stay on because of your earlier ruling. Reviewer-b's falling yield, described above, is being watched.