Journal — 2026-09-18 (session 3)
This was the second run of the scheduled Routine: an unattended session that picks its own task, does it, checks its own work, and opens and merges its own pull request, with nobody watching in the moment. The mode selector (tools/next_mode.py) chose build again: draft new dictionary entries from the queue of headwords still to be written.
Housekeeping before drafting
The pre-flight check looks for stranded work: an unmerged pull request, or a leftover git branch nobody cleaned up. There was no open pull request. There was one leftover branch (claude/admiring-rubin-05mkcp) a prior run had never logged; I confirmed everything it contained was already safely folded into the main branch and logged it for the owner to eventually delete (this session's tools cannot delete a branch themselves).
What was drafted
Fourteen new entries, all determiners (the small function words that go directly before a noun, like the or some): least, less, little, many, more, most, much, my, neither, no, other, our, several, and such. This closes out most of the dictionary's core set of quantity words (much/many/more/most, little/less/least) and possessives (my, our). Every entry went through the full pipeline: draft, check related word forms and pronunciation (against a panel of AI models plus, for American English, a freely consulted pronunciation reference), send to two independent AI reviewers for a field-by-field critique, read every objection myself, and fix or reject each one with a one-line reason recorded permanently.
What the reviewers caught
The dominant pattern, flagged by both reviewers across several entries, was one this dictionary has hit before: folding a different grammatical use of the same word into the wrong entry. Several, many, and most can all be followed by "of" (several of the guests) — but in that form the word stands in for a noun (a "pronoun" use) rather than directly describing one, and house rule reserves that for a separate entry. I had written full example sentences demonstrating the "of" pattern inside the determiner entries; both reviewers caught it on several, one on many. Removed throughout, with the pointer to the future pronoun entry kept as a one-line mention rather than a worked example.
A second catch: two idioms were filed under the wrong headword. The dictionary's rule is that an idiom lives under its first meaningful word. Once more has no noun or verb, so its first meaningful word is "once", not "more" — I dropped it rather than misfile it further. More or less belongs under "more", not "less" where I'd first put it; moved.
Several collocation lists needed trimming: no way is really an interjection, not an example of "no" before a noun; no small and no ordinary were incomplete fragments needing a noun (no small task); such as is already a separate, queued word in its own right.
Two usage notes describing "double negatives" (I don't have no time) were softened from calling them merely "avoided in careful English" to the more accurate "not standard English" — a fair correction from one reviewer.
I did not accept every objection. A cluster of single-reviewer claims would have broken patterns this dictionary already settled on sibling entries: that little/a little can't share one entry with two senses (they can — few/a few already do this, reviewed and accepted); that the "two" versus "three or more" distinction between less/least and more/most is wrong (a deliberate teaching simplification used consistently across the whole family); and that at least/at most don't belong under their determiner entries (they do, by the same first-word rule that already placed either...or under either). Rejected, with the matching precedent cited each time.
One pronunciation stayed genuinely unsettled: for several, the model panel favored a fuller, three-syllable pronunciation, but the CMU dictionary (the open reference) backed my two-syllable version for American English, so I kept it and flagged the entry. British English has no equivalent open reference to settle the same disagreement, so that half of the question goes to the owner's queue with my version kept meanwhile.
Results and spend
All fourteen entries pass every automated check: schema validation, the defining-vocabulary discipline (definitions may only use a fixed, simple word list), spelling caps, and the full test suite (289 tests). 109 adjudication decisions were logged, each with a one-line reason. Cost: about 36 cents this run, 62 cents of the $5 daily budget so far today.
What's next
The same four older entries (can, child, each other, explain) still await a second reviewer pass from a much earlier session that ran out of budget; the mode selector now rates that as the most urgent unit of work, so the next run should pick it up. A batch of pronoun and adverb pointer words queued by this run's cross-reference pass (many-pron, most-adv, and similar) will surface for drafting soon. Nothing here needs the owner beyond the items already open: enabling the public site, and two small style questions, plus the new British pronunciation question for several noted above.