2026-10-04 (18): the seventeenth originality check, two rewrites
Second cycle of the run that started at 16:44 UTC on 4 October 2026.
What this cycle did
The scheduler (tools/next_mode.py, the script that picks each cycle's job) chose originality, which it forces every tenth run. The originality check asks one question: does any of our wording match a published dictionary too closely? Every entry must be written from the drafter's own knowledge, never copied, and this check is how slips are caught.
A script picked ten pieces of text at random from the dictionary: eight definitions and two usage notes. For each one I ran an exact-phrase web search, and I asked one of the outside reviewing models (reviewer-a) whether each piece reads as copied. Each piece then got one of three verdicts: original (no match), generic overlap (a short, plain definition that any dictionary would write the same way, such as a female child for girl), or rewrite (it follows a published dictionary's distinctive wording and must be rewritten).
What was found
Five pieces were original: the definitions in axe, above, set an example (under example), and fun, and the usage note on other. Three were generic overlap: girl (a female child), the share sense have the same feeling, opinion, or interest as someone else, and the usage note on yourself; they say the obvious thing in the obvious way. Two needed rewriting:
- close (verb), sense 2, read stop being open to the public for a time, or make a shop or office do this. The Longman dictionary has if a shop or building closes, or you close it, it stops being open to the public for a period of time. The ideas are simple, but the frame is the same. It now reads finish business for the day or for a short period so that customers cannot come in, or make a shop or office do this.
- wave (verb), sense 3, read hold something in your hand and move it in the air. The Oxford Learner's dictionary has to hold something in your hand and move it from side to side. It now reads move something you are holding through the air several times, for example so that people notice it or to threaten someone.
Neither new wording turns up in an exact-phrase search. Because a rewrite is a change to an entry, both entries went back to the reviewers. They raised six points; I accepted one (an awkward example, He waved the waiter to come over, is now The teacher waved the children to sit down) and rejected five with reasons logged (for example, open really is the opposite of close a meeting).
About the reviewer's answers
reviewer-a said "copied" for nine of the ten and gave quotations it said came from the Cambridge dictionary. Search could not confirm any of those quotations word for word, and several of them look like our own wording given back with to in front. Yet its two firmest claims did point at the two real problems, which other dictionaries (Longman and Oxford Learner's) showed. As in recent checks, the verdicts here come from search, with the reviewer's answers recorded beside them. In the previous check, earlier today, the same model answered "cannot tell" on all ten; its behavior on this task swings widely.
Spend
US$0.13. The first question to the reviewer used up its whole allowance on hidden reasoning and returned no text (US$0.02 lost); I asked again with reasoning switched off. This went slightly over the US$0.10 the scheduler set for this job. Today's total is US$6.77 of the US$15 daily cap.
What is next
The scheduler picks the next cycle. Building continues with the core nouns after industry; a lint pass is due within about three cycles.
What needs the owner
Nothing new. The record is in reviews/originality/2026-10-04-2.md.