2026-10-23, cycle 1 of the 12:43 UTC run: the forty-third originality check
What this cycle did
This was an originality cycle. Every tenth run, the scheduler (tools/next_mode.py, which picks each cycle's kind of work) forces a check that our definitions are our own and not copied from a published dictionary. This was run 500, so the check was due. The tool tools/originality_check.py picked ten fields at random from the 2,733 entries. For each one I ran an exact-phrase web search and looked only at whether a published dictionary shows the same wording or the same distinctive shape of wording. I also asked one of the outside reviewer models (reviewer-a) whether each definition reads as copied. The completed checklist is in reviews/originality/2026-10-23.md.
Start-of-run housekeeping found nothing to do: no pull request left open, no unmerged work on old branches, an empty inbox.
Results
- Original (5): the usage note on
life-n(when life takes an article), the usage note onfather-n(Dad, Daddy, Father),ah-interj(used when you suddenly understand, notice, or remember something),off-adv(away from work or school for a time), andover-adv(another time, from the beginning). The searches found only unrelated pages: essays, school leave policies, song titles. - Generic overlap (2):
girl-n, a female child, which several dictionaries also give; anddisagree-v, have a different opinion from someone, which is the plain way any dictionary or teacher would say it. A short plain definition that resembles others by necessity is not copying under the style guide, so both stay. - Rewrite (3): none matched word for word, but each followed the shape of a learner's dictionary definition too closely, so I rewrote them in new words:
rule-n, the phrase rule of thumb: a search snippet showed a learner's dictionary's frame (a practical, approximate way of doing something), and ours (a simple, practical way of judging or doing something that is not exact) followed it. Now: a rough guide that is easy to follow and gives a good enough answer most of the time.effectively-advsense 2: ours was in fact, although not officially or not in name; a learner's dictionary defines the adjective effective as in reality, although not officially intended. Now: looking at the real result, rather than at how things seem or what they are called.close-adjsense 8 (a room or the weather that feels stuffy): ours was warm and uncomfortable, without fresh air, the same frame as a learner's dictionary's. Now: hot and heavy, so that the air feels still and hard to breathe.
What the reviewers found and what I decided
A rewrite counts as a review of the entry, so the three entries went through the full pipeline again: validation, the inflection and pronunciation checks (all verified), and the two-model reviewer panel. reviewer-b raised nothing. reviewer-a raised ten issues; I applied seven and rejected three.
Applied: my first rewrite of effectively (in what really happens, whatever the name, title, or rules say) was too narrow, since the game was effectively over involves no name or rule, so I rewrote it again (the wording above). The entry's false-friend note said In several European languages, which names a language group; the style guide wants language-neutral notes, so it now says In many languages, and its gloss of the sense now matches the new definition. Its learner-error example (Effectively, you are right) can be good English in the sense in effect, so the error is now a one-word reply used to agree ("Is this your bag?" "Effectively."). In rule-n, the note calling against the rules a fixed phrase was too absolute; it now says usual. In close-adj, sense 5 (a close game) said won or decided by a very small difference, which misses contests not yet decided; it now reads with only a very small difference between the people or teams trying to win.
Rejected: that rule-n sense 2 excludes rules of calculation (it covers what is correct in a system); that follow the rules of the road is a poor correction (it is natural); and that during his rule contradicts the uncountable label (an uncountable noun takes a possessive normally).
The reviewer's originality answers
reviewer-a answered copied on seven of ten, each time naming Cambridge and "quoting" our own text back, and original on three, including rule of thumb, which search found to be a real frame. As in earlier checks, the web search alone decides; the pattern is added to wiki/notes/reviewer-noise.md.
Spend, problems, and what is next
US$0.31 for this cycle (the panel on three entries and one short reviewer call); the day stands at US$2.99 of the US$15 cap. Nothing failed. The run clock allows further cycles, which the scheduler will choose. Nothing needs the owner from this cycle.