Run report, 2026-10-30 (second run of the day, started 06:44 UTC)

Cycle 1: build

What this cycle did. The selector (the script that picks each cycle's kind of work from the project's state) chose build: write new entries from the queue (the ordered list of words still to be written). It allowed up to 12 entries and US$3 of paid model calls. I wrote twelve common ("band 2") verbs, the next ones in the queue order: drain, edit, embarrass, emerge, emphasize, enable, ensure, equal, erase, establish, estimate, and evaluate. Every one went through the full pipeline: the schema check, generated verb forms, the pronunciation panel (three models plus an open pronouncing dictionary, voting on my transcriptions), both reviewers on every field, my reading of every issue, and the cross-reference check. The dictionary now has 3,341 entries, all reviewed.

Some choices in these entries:

What the reviewers found and what I decided. The two reviewers raised 63 issues, 3 of them blocking (they must be settled before an entry is published). I applied 52 and rejected 11.

Pronunciation. The panel disputed my British transcription of ensure, ɪnˈʃɔː: only 1 of 3 votes agreed, and the others gave ɪnˈʃʊə. Both are real British pronunciations. I kept ɪnˈʃɔː to match the existing entry sure, added ɪnˈʃʊə as a British variant, flagged the entry, and wrote a line for you in the curator file. This follows how sure (adverb) was handled on 2026-10-17.

What did not work. A word that one entry names as a synonym should list the first entry back, at the matching sense; these mirror links are called back-links. The cross-reference tool did not write seven of them because it could not tell which sense to attach them to, so I placed them by hand. Two of them (on tire and appear) I found only by checking myself, because the tool's report was cut short in my view. I also dropped change as a synonym of the computer sense of edit, since it was too loose to mirror.

Spend. US$0.40 this cycle, for the pronunciation panel and reviewer B. Reviewer A is billed to your subscription and costs nothing here. That makes US$1.43 of the US$15 daily cap so far today.

What is next. More band-2 verbs from the queue, in its order. A lint pass (the periodic whole-dictionary check) is due next run.

For the owner. One new curator line: the British pronunciation of ensure.

Cycle 2: lint

What this cycle did. The selector forced a lint pass. This is the whole-dictionary check, due every ten runs, and ten runs had passed since the last one on 2026-10-29. A lint pass runs the mechanical checks over all 3,341 entries and then reads the project's own records for contradictions and gaps. It writes no new entries.

Mechanical checks. The word caps on the rule files and the wiki held, every link between pages resolved, and the gates passed: the schema check, the defining-vocabulary check and the cross-reference check. Six spent claim files were removed. A claim file records which words a run is writing, so that two runs do not take the same word.

Cross-references. When one entry names another as a synonym or a word to compare, the other entry should point back at the matching sense. The tool wrote 7 of these back-links itself. It could not choose a sense for 18 more, so I placed 13 by hand. Two came from this run's first cycle (ensure back to see, and embarrass back to show up). Most of the rest came from the verbs built in the previous run (declare, defeat, delay, detect, determine, disturb). The other 5 are the same ones listed before: in each case the forward reference names no real sense of the target. They stay in the cross-reference audit that your ruling C1 asked for.

Band vote. 81 queue rows had been added by tools since the last vote with no frequency band. The three word-list models voted on them for US$0.013. Two rows split by more than one band, ma'am (noun) and underestimate; I settled both at the middle vote, band 3. All 81 rows are now in the queue with bands.

Reviewer precision. Each lint measures how often each reviewer's issues, by kind, turn out to be right. Reviewer B's phrase issues (idioms and fixed phrases) have now reached twenty decisions with only 4 applied, or 20 percent. The standing rule switches off any kind of issue under 30 percent once it has twenty or more decisions, so I switched this one off for reviewer B. Most of the rejected issues asked for an idiom that the house rules file under a different keyword, or for a phrasal verb to be added as a phrase. Reviewer A still checks phrases. Reviewer B's spelling-or-format issues are also under the line (26 percent), but they stay live because you ruled that they should.

Reports kept for later review cycles. These are unchanged in kind:

Records. The wiki index matches the pages, the log has an entry for every day, both open questions are still open and still apply, and the curator file has no duplicate lines.

Spend. US$0.013 this cycle; US$1.44 of the US$15 daily cap today.

What is next. Building band-2 verbs in queue order. The next lint is due in ten runs.

For the owner. Nothing new beyond the ensure pronunciation line from cycle 1. Reviewer B no longer raises phrase issues. If you would rather keep that check on, a one-line ruling restores it.

Cycle 3: review

What this cycle did. The selector chose review. There were no draft entries, so the cycle continued the coverage pass. This is your ruling B1: a second reading of the most important band-1 entries, most senses first, to add common uses that the first draft missed. This cycle read four entries with seven senses each: home (noun), hole (noun), force (noun), and fine (adjective). 105 entries have now had this reading.

What I added.

What the reviewers found and what I decided. The two reviewers raised 63 issues, 6 of them blocking (they must be settled before publication). I applied 58 and rejected 5.

Spend. US$0.27 this cycle; US$1.71 of the US$15 daily cap today.

What is next. The coverage pass continues with fill (verb), then the other seven-sense entries in the same order.

For the owner. Nothing new.

All journal entries