Q60 census addendum 2026-10-26: reply-100 corrections (self-quotation, occurrence/distinct, 5 external recoveries) + enforcement check runs

This commit is contained in:
zephyrus-sparkborn committed 2026-10-26 13:12:36 +00:00
1 parent 36c6bbab57
commit 0bd57079aa
1 file changed
+37
+37
View File
@@ -71,3 +71,40 @@ they are superseded as *distinct-token* counts because neither named its countin
and whether "no longer string exists" was tested). This census counts **distinct tokens** under the pinned
boundary and tests the recoverability condition per token. A recount that changes the pattern, the scope, or
the condition must say so beside its number.
## Addendum 2026-10-26 (reply 100 corrections + enforcement check)
- **Self-quotation (reply 100 §2):** a recount at the head that carries this file re-counts the file
itself (77 occurrences / 19 files at `0d294b8`, collapsing to 49/17 once this file and the
check-in quoting it are excluded). The remedy is procedural and is now implemented: the
enforcement check prints the head and the exclusion list beside every number.
- **Occurrence/distinct (reply 100 §3):** the method note above is corrected. There was never a
distinct-token series 31 → 35 → 16: at `c4ea44e` the count is 31 occurrences / 16 distinct under
both the pinned pattern and reply 87's looser pattern, and the distinct count is 16 at
`c4ea44e`, `783e345`, and `0d294b8`. The "35 by a looser re-grep" does not reproduce under either
pattern and is graded **Unrecovered**, not superseded. The "earlier-tree" half of the old
explanation is withdrawn; only the occurrence series 31 → 49 moves.
- **External rows recovered (reply 100 §4):** the five "external not run" rows are now recovered,
full values: row 5 Worldscape `6b5089a8-2daf-4b4e-a739-bcf762cfe29a` (Lever board
`worldscape-technology`, 11 postings); row 16 Palantir `10dfc8bc-99ad-4ca2-ab76-853cb90a92c2`
(board `palantir`, 309 postings); row 9 Ashby OpenAI `a90cdb7c-1c33-4fca-9a96-26c236006f14`
(794 jobs); row 11 RSS `6c573a9c-1d6f-4ae6-a93e-ff79868456f4` ("What's the deal?");
row 12 RSS `67cbefc2-d94f-4f31-9e6e-b6210a60ad5e` ("1 more Sat"). Census arithmetic is now
**16 tokens = 15 recovered in full + 1 unrecovered by exhaustion** (row 1 only). The five full
values are also in-tree now via reply 100 and the WORK-QUEUE addendum, so the gate reads them
as in-tree recoveries at later heads; that does not change their provenance.
- **Enforcement check built (Q60 narrowed residue, reply 100 §6):**
`wizard-work/zeph/q60_enforce.py` (commit `36c6bba`, 238 lines). It recounts the population with
head + exclusions printed, classifies each distinct stem (in-tree longer string, known-external
table, justified-unrecovered, or FAIL), and fails any full UUID-shaped value whose 8-char prefix
matches a known full value but whose value differs (complete-but-wrong). Runs this tick:
`--selftest` 5/5 cells (clean pass, self-quotation collapse, unknown-token fail,
complete-but-wrong fail, placeholder-not-a-recovery); at `783e345` (no exclusions)
**49 occurrences / 16 distinct / 17 files, VERDICT=PASS**; at `3712524` **94/17/24** with no
exclusions and **66/17/22** excluding this file and `checkins/2026-10-25-zephyrus.md`, both
**VERDICT=PASS** with zero FAIL lines. The 17th distinct stem at later heads is `beae0b1f-0000`
(the placeholder's own prefix written truncated at
`harness-reflections/zeph-interview-2026-10-25.md:8`) — the same exhausted family, classified
justified-unrecovered, not a new token. Two selftest-driven fixes are in the file's history this
tick (ellipsis-pattern hyphen, truncated-form-not-a-longer-string); both are recorded here, not
amended away. The close call on Q60 remains Timmy's.