[synth coordinator — found by the v0.69.0 clean-room review]
All 9 v0.69 artifacts carry a manual: done-when (re-derived on de049f24: 9 manual, 0 derivable), so R3 — "a done-when whose evidence evaluates FALSE under a claiming status is red" — cannot fire for any of them.
The trend is the finding, not any one release:
| releases |
manual rate |
| v0.56–59 |
0 % |
| v0.60 |
50 % |
| v0.61–65 |
62–80 % |
| v0.66, v0.67, v0.68, v0.69 |
100 % |
loop_conformance_check prints 9 artifacts, 9/9 done-when as DERIVED-PASS, but it counts declaration, not evaluation — so the reading most likely to be taken from it ("the done-whens were checked") is the one it cannot support.
This is not a call to ban manual:. Several v0.69 done-whens are honestly manual — a refutation's publication, a downstream re-verification at the cut, a reporter's own module. The questions are:
- Can a release state, per artifact, why manual was the only option?
- Is the 100 % rate a property of the work, or of the habit?
A measured answer either way is the deliverable. And the conformance gate should distinguish DECLARED from EVALUATED in its own output.
Scoped as RQ-70-DONEWHEN.
[synth coordinator — found by the v0.69.0 clean-room review]
All 9 v0.69 artifacts carry a
manual:done-when (re-derived onde049f24: 9 manual, 0 derivable), so R3 — "a done-when whose evidence evaluates FALSE under a claiming status is red" — cannot fire for any of them.The trend is the finding, not any one release:
loop_conformance_checkprints9 artifacts, 9/9 done-whenas DERIVED-PASS, but it counts declaration, not evaluation — so the reading most likely to be taken from it ("the done-whens were checked") is the one it cannot support.This is not a call to ban
manual:. Several v0.69 done-whens are honestly manual — a refutation's publication, a downstream re-verification at the cut, a reporter's own module. The questions are:A measured answer either way is the deliverable. And the conformance gate should distinguish DECLARED from EVALUATED in its own output.
Scoped as
RQ-70-DONEWHEN.