Journal — 2026-09-19 (run 8 of the day)
This was the tenth scheduled run of the dictionary project overall, and it ran in build mode — drafting new dictionary entries. The scheduler picks build when it has accumulated the most "debt" among the eligible modes; today that meant new content was more overdue than a review pass, a site rebuild, or another originality check.
Start-of-run checks. No pull request was left open from a previous run, no abandoned work branches existed on GitHub, and the inbox (where the owner can drop instructions) was empty. The checkout was current with the main branch.
What got drafted. The queue tool handed back 20 pronoun headwords. Rather than draft all 20 thinly, the run kept 13 that form a tight, coherent group — the personal pronouns I, me, she, him, his, hers, mine, and ours (subject, object, and "independent possessive" forms, the kind used without a following noun, as in this book is mine), plus the five reflexive pronouns myself, himself, herself, itself, and ourselves (the kind that refer back to the subject, as in I hurt myself). The other seven (neither, nobody, none, nothing, other, some, somebody) were put back in the queue for a future run rather than rushed.
The review catch worth noting. Every new entry goes through two independent AI reviewers before publication. Across four of the five reflexive entries, both reviewers caught the same real mistake: a usage note claiming that a reflexive pronoun like myself is used "only when the subject and the object of the same clause are the same person." That's true for the reflexive sense (I cut myself) but false for the entry's other, emphatic sense (I myself don't believe it — no object there at all). The note was rewritten in each entry to apply the restriction only where it belongs. A similar self-contradiction turned up in the itself entry, which claimed English "normally" drops the reflexive pronoun with verbs like open and close — directly contradicted by the entry's own example, the machine turns itself off. That claim was softened to match reality rather than deleted outright. The she entry's list of common learner mistakes said "she is used only for people, and sometimes animals, never for a thing," which flatly contradicted the entry's own second sense (the old-fashioned use of she for a ship or country). Fixed the same way: qualified rather than removed, since the underlying teaching point (don't call an ordinary object "she") is still sound.
Two entries also had definitions using words outside the dictionary's approved list of roughly 2,400 "defining vocabulary" words that any definition is allowed to use, so a learner is never sent to a harder word to understand an easier one; those got reworded. One pronunciation transcription (me, British) was missing a vowel-length mark it should have had, caught by a reviewer and fixed.
A pronunciation dispute left open. For ourselves, the American pronunciation panel of AI models and a reference dictionary both leaned toward a fuller, two-syllable rendering of the first part, disagreeing with the drafter's shorter, one-syllable form. Unlike a similar disagreement on ours (where the reference dictionary sided with the drafter, settling it), nothing here settled the question, so the entry is flagged as disputed and a line was added to the curator queue for a human to weigh in eventually; the drafter's form stays in place meanwhile.
Numbers. 13 new entries, all now published (reviewed status). 34 issues raised across both reviewers, 18 serious enough to require a decision; 30 fixes applied, 4 reviewer objections rejected as wrong (with reasons logged, including two that matched wording already approved in sibling entries like he and her). Spend: US$0.34 of today's $5 cap (US$2.04 spent today in total).
What's next. Two large batches remain queued for a future review pass: 23 large, multi-sense entries still need a formatting-marks update, and 120 entries have had only one review round instead of two. A lint pass is due in about four more runs. Nothing here needs the owner's decision beyond the three standing open items already on record (GitHub Pages activation, whether to split some grammar entries by part of speech, and the visual style of the formatting marks) plus the new pronunciation question on ourselves.