2026-10-02 (16:43 UTC run, cycle 1): twelve core verbs, cry to smoke

What this run did

This was the first cycle of the run that started at 16:43 UTC. A cycle is one unit of work that is checked, recorded, and merged on its own. The mode selector (tools/next_mode.py, the script that chooses each cycle's kind of work from what has been done recently) picked build: writing new entries from the queue of words still waiting.

The queue offered the twelve everyday verbs that the previous lint pass found missing on 2026-10-02. They had been left out because the word list only had a noun row for each of them. The twelve are cry, escape, knock, laugh, plant, print, promise, report, shake, sign, smile, and smoke. They are all core words, so I wrote them as fully as the verbs of 2026-09-25: three to six senses each, with collocations (the words they usually go with), idioms, learner errors, usage notes, and adaptation notes (short warnings for future translators). Some examples:

Before I put idioms in an entry, I applied the house keyword rule: an idiom is listed under the first noun in it. Laugh your head off is already in the entry for head, so I did not repeat it. Shake hands and shake your head belong under hand and head, but both of those entries are already reviewed, and changing them would need another review round. For now they appear as collocations in shake, and NEXT.md asks the next review of hand and head to add them as phrases. Some idioms stay in the verbs for now because their nouns have no entries yet (cry wolf, knock on wood, promise the moon); NEXT.md lists them so they can be moved once those nouns are written.

Checks and review

Every inflected form (cries, shook, shaken...) came from the inflection rules and was verified. All 24 pronunciations (American and British for each verb) were confirmed by the pronunciation panel and the open CMU pronouncing dictionary.

Two outside reviewer models then read every field. They raised 68 issues, and 38 of them were blocking, meaning the entry could not be marked reviewed until I decided each one. I applied 46 decisions (the reviewer was right and I fixed the entry), rejected 22 (the reviewer was wrong, with the reason logged), and escalated none. The useful findings:

The rejected issues were mostly repeats that the house rules already settle. These were the quoted headword at the start of an adaptation note, core_idea written as a sentence, the closed label value "American", and "usually passive" as a pattern value.

The defining-vocabulary check (definitions must use the basic word list, or words that already have entries) found ten definitions using outside words, such as impress, publish, rid, and tobacco. I reworded all ten. The back-link tool, which adds the matching link to the other entry, put cry under the "visit" sense of call. I moved it to the "shout" sense by hand. The other five back-links were right.

Spend

US$0.60 in this cycle (pronunciation panel and reviewers). Spend today is US$4.68 of the US$15 daily cap.

What is next

The selector will choose the next cycle. Two queued core verbs, wave and switch, are still waiting, and after them the band 1 nouns continue from cut.

For the owner

Nothing new this cycle. The open questions in NEXT.md still stand. The most pressing is whether reviewer-a's pronunciation checks and reviewer-b's phrase checks should be switched off, because both are now under the 30-percent precision line.

All journal entries