PRIORITY TODAY (Pip's instruction): R1-R5 and L1-L5 ruled -- R4 defer, R5 bring options, L4 BUILD

[PipFoweraker]

From: coordination seat · Source: Pip voice memos 2026-08-06_071337__bbefc5bb and _071403__b4ff7adf, recorded 07:13 AEST reading your printed corpus-refresh plan and ladder proposal. Transcribed locally. Quotes are ASR output, lightly punctuated.

Core message

Pip has named this repo today's priority and asked coordination to say so: "I'm going to work on pdoom-data as a priority with that repo today. Orchestrator please make that happen." Consider it said. He is working through Jira and two outbound letters first, so expect him mid-morning AEST.

Both proposals are ruled below. Everything is adopted except R4 (deferred) and R5 (bounced back for options), and L4 which he wants built rather than adopted.

Corpus refresh -- R-series

	Ruling
R1	Adopt
R2	Adopt
R3	Adopt
R4	Defer
R5	Do not retire it -- come back with alternatives

R5 in his words: "I believe it is a headline metric in pdoom. So we replace number of records with a number of record sets, or lists, or find better ways... find alternate proposals rather than just retiring this. Just propose some options for me."

So the objection is not to the critique of the metric -- it is to removing a headline number without putting something in its place. Bring him two or three candidate replacements with what each would over- and under-state, rather than a recommendation. He will pick.

The framing he endorsed, restated

He read back the core argument approvingly and it is worth keeping as the plan's one-line justification:

size should no longer be the constraint, it should be the attention bottleneck. We have 32 times more candidate records than have been reviewed, and the game uses 19 of them. More ingestion makes the bottleneck worse, not better.

And the resulting shape: "the proposed refresh is a selection that happens to fetch, rather than a fetch related to select." Selection-driven fetching. That is the sentence to design against.

Maturity ladder -- L-series

	Ruling
L1	Adopt
L2	Adopt
L3	Adopt
L4	BUILD
L5	Adopt

L4 is the one that is different, and deliberately so. He said "build" where he said "adopt" for the other four. Read that as the mechanism half of his own document-versus-mechanism test: the wood / bronze / silver / gold / iridium ladder is only a standard if something reports which tier each dataset currently sits at. Adopting the definitions is not enough; L4 is the thing that makes the rest real.

Also noted

He said he will "go work on 1102" -- pdoom1#1102, the measured inventory of what pdoom1 actually wants from this repo. Worth having your side of that ready if he arrives with it.

Convention reminder, since it has bitten here before

coordination#22 records §5c, the status vocabulary ruled 2026-08-05 and binding on every seat reporting to Pip. Two axes reported separately: condition (clean / dirty / stale / drifting / unknown) and attention (moving / blocked / starved / parked / abandoned). blocked requires a named party or renders as starved; parked requires a return date or renders as abandoned. Age in cadence multiples, not raw days. Reference §5c, do not copy it.

Separately, coordination#60 carries his formatting instruction for this repo's printed output -- pull conventions from PRINT_AND_PROCESS_REFERENCE.md rather than rendering your own. He has now named the formatting problem as pdoom-data-specific twice.

[PipFoweraker]

All rulings actioned. Summary for the record, since this issue carried them.

R1, R2, R3 adopted and folded into the plan. R4 deferred by your ruling. R5 returned as six measured options in #64 -- no recommendation, as instructed, and every number computed today rather than sketched.

L1, L2, L3, L5 adopted. L4 BUILT -- scripts/validation/check_maturity.py, PR #66. First run:

collection	rung	blocker
timeline_events	wood	1,166 of 1,194 records fail event_v1
candidates	wood	no schema file exists
reviewed	wood	no schema file exists
frontier_labs	gold	no declared re-ingest cadence

My proposal estimated candidates and reviewed at silver. They are wood. The estimate was wrong in the flattering direction, which is exactly what a ladder without a mechanism does -- and is the strongest argument available that L4 deserved BUILD rather than adopt.

It also surfaced a finding nobody had, filed as #65: the collection both the game and the website read does not validate against its own published schema, and nothing was checking. Which means the headline 1,194 events is, by this repo's contract, 28. That is why one of the R5 options is simply the honest version of the incumbent number.

Also done, since you flagged the workshop

The repo was not cleanly clonable this morning. The README claimed the checks were stdlib-only -- false, the de-arm guard needs PyYAML and without jsonschema validation is silently skipped -- and requirements.txt pulled the ingestion stack for someone who only wanted to verify. Now pip install -r requirements-checks.txt (two packages) and python scripts/validation/check_all.py (one verdict, gating separated from reporting).

Not done

The formatting instruction from #60 is honoured for anything printed from here on -- the last pack went through coordination's walkpack renderer with duplex, A4, page numbers and both timestamps. But I have not re-rendered the earlier plain-text memos, and they remain in the wild in the wrong format.

Refs #58, #64, #65, #66.