Journal — 2026-09-20 (run 7)
This was the twentieth 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) picked review — a second reading of existing entries — because no new entries were waiting to be drafted and review had accumulated the most "debt" among the modes the selector weighs.
Housekeeping before the review
The pre-flight check looks for stranded work: an unmerged pull request from a previous run, or a leftover git branch nobody cleaned up. There was neither — the previous (lint) run had already merged cleanly. The inbox, where the owner would leave instructions, was empty, and the checkout was already current with the shared history, so I moved straight to the review unit.
Clearing the inline-markup backlog
On 2026-09-18 the owner ruled that the dictionary's prose needs two typographic marks: bold around a word discussed as a word itself (the word all goes before a plural noun), and italics around a phrase quoted as an example of use (it rained all day). Entries written before that ruling needed a second pass by hand, field by field, to add these marks without changing any wording — never done by a script or a find-and-replace, since judging "is this word being named, or just used" needs a reader's understanding of the sentence. Earlier runs had been chipping away at this list, smallest entries first. This run finished it: the 15 entries still waiting — the dictionary's longest, most heavily used words, house, shit, bank, color, water, good, thing, make, see, head, go, get, way, time, take — all got the treatment in this one run, closing a backlog that had been open for two days.
Because the entries are independent files, I worked them as fifteen separate, simultaneous passes rather than one after another, each one reading every prose field of its entry (definitions, usage notes, example glosses, translator's notes, and so on), checking it against the style guide's rule and against three already-marked "model" entries for precedent, and adding marks only where they belonged — most fields, it turns out, need none. Roughly 700 bold marks and 450 italic marks went in across the batch. None of the underlying wording changed; every file was checked afterward by the dictionary's own validator, which enforces that marks are correctly paired and never misused, and every one came back clean.
One thing surfaced that isn't a markup problem: an example sentence under the entry for time, illustrating the phrase "kill time," instead says "I had an hour to kill" — a natural variation that doesn't literally contain the word "time," which trips a mechanical check meant to catch missing headwords. That check is doing its job; the sentence itself may be fine or may want a small rewrite. I left it as a note for a future review run rather than rewording it myself, since content changes are outside this run's job.
Results and spend
All fifteen entries now carry zero style-guide flags — the entire inline-markup conversion project, seeded on 2026-09-18, is complete. Every project check passes: schema validation, formatting caps, links, the full 305-test suite, vocabulary discipline, and cross-references (728 "missing target" warnings remain, but those are the normal, expected gaps where an entry references a word the dictionary hasn't written yet — a running list, not an error). Spend: essentially nothing, since hand-marking prose needs no paid model calls; the daily budget stands at $1.26 of $5.00, unchanged from earlier today.
What's next
With the markup backlog closed, the review queue is now empty of both drafts and flagged entries; the next time the selector picks review, the unit will be a second independent-reviewer pass on entries that have only had one so far (173 of them). Otherwise the queue is unchanged: seven prepositions ready to be drafted, and a growing list of over 400 words the dictionary already refers to but hasn't defined yet. The next lint pass is due in about four more runs; the next originality spot-check is due around run 30. Nothing here needs the owner beyond the three items already open: turning on the public website, and two small style questions (whether four grammar-heavy words should be split into separate entries by part of speech, and what color or weight the dictionary's bold/italic word-marking should use).