2026-10-12, run at 08:44 UTC, cycle 1: the sixty-seventh lint pass

What this cycle did

This cycle was a lint pass: a scheduled check of the dictionary and its knowledge base for problems that a single entry's review cannot see. The scheduler forces one after five runs without one, and the previous run (04:43 UTC) was four cycles of building and reviewing, so this one was due. A lint pass costs nothing, because it makes no calls to paid models.

The pass has a mechanical half (scripts that check every file) and a judgmental half (reading pages and entries against each other). Nothing serious turned up. Fifteen entries changed, all in their cross-references; two curator lines were moved to the right place.

The mechanical checks

Reviewer precision

tools/metrics.py --precision reports, for each reviewer and each kind of issue, how often I applied that reviewer's point. A kind of issue on which a reviewer 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 72 percent of the time overall, reviewer-b 53 percent. Since the last pass, reviewer-b's points were applied 20 times out of 35 over four cycles. The one live pair under the line, reviewer-b on spelling and format, stays on by the owner's ruling of 2026-10-05. Details are in wiki/notes/reviewer-precision.md.

The judgmental checks

Spend

None. The day's spend stands at US$5.58 of the US$15 cap.

What is next

The next lint pass is due in about five cycles. The run may start more cycles if the clock allows.

For the owner

Two pronunciation questions from this morning are now in the open part of the curator file: British military (the checking panel split three ways between MIL-i-tree, a reduced middle, and four syllables) and part-time (the panel's stress changed between calls). The three old claude/wonderful-bardeen-* branches can be deleted; this tool set cannot delete branches.

All journal entries