Journal, 2026-11-06 (run fwhuo1)

Cycle 1: build

What the cycle did. The mode selector (tools/next_mode.py, the script that picks each cycle's job from what is most overdue) chose build: writing new entries from the queue, the ordered list of words still to be added. The cap for a cycle is twelve new entries, and this cycle wrote all twelve, all nouns in frequency band 2 (common words just outside the core vocabulary): estate, estimate, ethics, ethnicity, European, evaluation, ex, ex-wife, examination, excess, executive, and exhibit. Each went through the full pipeline: schema check, inflections from the spelling rules, a pronunciation check by a panel of models plus the open CMU pronouncing dictionary, two independent model reviewers reading every field, my own decision on each point they raised, and the closing checks. All twelve are now marked reviewed and will appear on the site at the next build.

Some content worth knowing about:

What the reviewers found and what I decided. The two reviewers raised 15 issues, none of them blocking. I applied 14 and rejected 1. The applied ones were small but real: an example for the British housing sense used the American word highway; one evaluation example belonged to the worker-review sense, not the general one; a medical examination was filed under the wrong collocation type; excess in its first sense is both countable and uncountable (an excess of food, any excess); the party or union executive is mainly British and now has that label; and ethnicity said the same thing about nationality in two fields, which breaks the one-fact-one-field rule. The rejection: reviewer B asked for the plural ex-wives to be marked by hand in an example, but the house rule is that the site marks every listed inflection itself.

Separately, the vocabulary check found ten words in definitions that are outside the defining vocabulary and have no entry yet (property, culture, tradition, gallery, insurance, union, philosophy, grapes, extreme). I rewrote those definitions in plainer words; the vocabulary check now passes.

Pronunciation. 23 of the 24 transcriptions were verified. The American excess (stress on the second part) got only one of four panel votes, though the CMU dictionary agrees with it. Both stress patterns are heard, so I kept the second-syllable stress as the main form, listed the first-syllable one as a variant, flagged the entry pronunciation-disputed, and added a line to reviews/needs_curator.txt for a session with you.

Cross-references. The back-link tool could not choose a sense in two older entries, so I placed them by hand: executive is now a "compare" on manager (sense 1), and evaluation a synonym on assessment (sense 1). One gap turned up: the existing race entry has only the competition senses. The people-group sense comes from a different origin, so by the house rule it needs its own homograph entry, race-n-2. ethnicity already points to that slug, and NEXT.md asks the next build to hand-claim it.

Spend. US$0.33 for the cycle. That brings today's total to US$2.52 of the US$15 daily cap.

What did not work. Nothing failed. One surprise: several everyday words (property, culture, gallery) still have no entries. They are on the queue and will come up in order.

What is next. The next cycle will be whatever the selector picks: most likely more building from the queue, after exhibit.

What needs the owner. Nothing new beyond the excess pronunciation line in the curator file.

Cycle 2: review

What the cycle did. The mode selector chose review: rereading existing entries to raise their quality. No entries were waiting as drafts, so the cycle carried on with the B1 coverage pass. That is the owner's ruling to reread the most important everyday words, those with the most senses first, and add whatever a learner would miss. This cycle took the next four of the 111 five-sense band-1 entries not reread since 2026-10-24: bring, burn, busy, and cheap. I read each one against my own knowledge of the word before sending it to the two reviewers.

What I added before the review.

What the reviewers found and what I decided. The reviewers raised 43 issues, 3 of them blocking. I applied 28 and rejected 15.

One process slip, corrected. I first logged the ten brought rejections against the same issue ten times. I removed those ten lines before committing and logged each rejection against its own issue. The decision ledger in this pull request is correct.

Other. The adverb cheap (I got it cheap) has no entry, so I queued it.

Spend. US$0.22 for the cycle. That brings today's total to US$2.74 of US$15.

What is next. The coverage pass continues with check, cold, color, and comfortable.

What needs the owner. Nothing new.

Cycle 3: build

What the cycle did. The selector chose build again. I wrote eight new entries, fewer than the twelve allowed, to keep each one full and to leave room for checking. The first, race-n-2, was the gap found in cycle 1: the existing race entry covers only contests of speed, and the sense of a group of people comes from a different origin, so the house rule gives it a separate entry, a homograph numbered 2. The other seven came from the queue in order: exhibition, existence, expansion, expectation, expense, expertise, and explosion.

Points of interest:

What the reviewers found and what I decided. The reviewers raised 17 issues, 3 of them blocking, and I applied all 17.

Something to watch. The script that stamps each new entry with its frequency band and defining-vocabulary flag (it reads them from the queue) gave race-n-2 the band and flag of the old race-n queue row: band 1, and in the defining vocabulary. That is plausible, since the defining vocabulary probably needs this sense of race, but no band vote has looked at it. House rules say bands come only from the vote, so I left the value as stamped and noted it in the log for the next band vote.

Cross-references. The back-link tool could not choose a sense in cost, so I placed the link to expense in sense 1 (price) by hand.

Spend. US$0.24 for the cycle. That brings today's total to US$2.98 of US$15.

Why the run ends here. The run clock allowed a fourth cycle. I stopped after three because a long run uses up working context, and the routine asks for the run to end before later entries come out thinner than the early ones.

What is next. The selector will choose. Building goes on after explosion. The coverage pass goes on at check, cold, color, and comfortable.

What needs the owner. The American excess pronunciation line from cycle 1, in reviews/needs_curator.txt. Nothing else.

All journal entries