Repository navigation
designations: record the 39 phase-3 day-zero designations (likelihood lifecycle) #83
Description
Overview
Phase 3 of epic likelihood-lifecycle: record the human's day-zero designations. On 2026-10-31 the human approved all 39 designatable likelihood × hardware-class groups re-derived on the day-zero catalogue (autolens_profiling #496 head c218a428), and held datacube/delaunay gpu_a100 as provisional because it does not separate (1.03× at sma).
Heart RED override invoked by the human 2026-10-31 (release-integrate failure, unrelated to Pulse decision records).
Plan
- Write 39
decisions/<id>.mdand 39kind: designationrows indecisions.yaml, dated 2026-10-31, from Appendix B as approved. - Each record gives the decision maker and source, then the per-instrument evidence: setups, projected calls/s/node, per call, separation against tolerance, and parity.
- Each record also gives
basis: projectedin prose (η = 1.0, declared and unmeasured), the rule, secondaries, untested branches and caveats. - Name the datacube A100 hold in the PR. It is not recorded as a designation.
- Generate the records from the catalogue with the reviewed generator, never by hand.
pyauto-pulse checkand the test suite must pass. - Out of scope: the autolens_profiling producer side (
class: optimal+decision_id, retiringprovisional_designations.json). That is the next task, after #496 merges. - Tier: judge — merge mode: human /prm
Detailed implementation plan
Affected Repositories
- PyAutoPulse (primary)
Branch Survey
| Repository | Current Branch | Dirty? |
|---|---|---|
| ./organs/PyAutoPulse | main | clean (untracked tmp/) |
Suggested branch: feature/likelihood-lifecycle-p3-designation-records
Implementation Steps
- Worktree
~/Code/PyAutoLabs-wt/likelihood-lifecycle-p3-designation-records/PyAutoPulse. - The generator reads autolens_profiling
dashboard/catalogue.jsonat #496 headc218a428(the ranking fromdesignate.py; entries fromappendix_b.py). It writesdecisions/<id>.md(template of the existing records: title, Date, Decision maker, Decision source, Campaign, Evidence, Decision, Rule, Caveats) and appends the index rows. - Validate:
pulse.decisions.load(),pyauto-pulse check --offline,pytest,ruff. - Add a short
decisions/README.mdnote listing the day-zero designation set, if the README keeps such lists. - PR → human
/prm.
Key Files
decisions.yaml,decisions/<id>.md(39 new)- Mind
draft/research/autolens_profiling/likelihood_lifecycle_p3_designation_proposals.md(Appendix B, the approved draft)
Original Prompt
Click to expand starting prompt
Likelihood lifecycle phase 3: record the 39 day-zero designations in PyAutoPulse
Type: research
Target: PyAutoPulse
Repos:
- PyAutoPulse
Themes: - profiling
Difficulty: medium
Autonomy: supervised
Priority: high
Status: formalised
Consequence: judge
Witness: 39 decisions/.md records and 39kind: designationdecisions.yaml rows, each matching Appendix B of likelihood_lifecycle_p3_designation_proposals.md as approved; pulse.decisions.load() passes; pyauto-pulse check and the Pulse test suite pass
Review-minutes: 30
Unattended: never
Epic: likelihood-lifecycle
Phase: 3
Parent: draft/research/autolens_profiling/likelihood_lifecycle_epic.md
Filed: 2026-10-31
Request (human, 2026-10-31, verbatim)
"contonue do those steps" — then, via AskUserQuestion: "Approve all 39 (Recommended)" (record the 39 designatable groups as drafted) and "Hold, keep provisional (Recommended)" (datacube/delaunay gpu_a100 does not separate); "Override + create issue" (Heart RED).
What
Record Appendix B (re-derived 2026-10-31 on autolens_profiling #496 head c218a428) as 39 Pulse designation decisions, date 2026-10-31, decision maker James Nightingale. Each canonical record states the per-instrument evidence (setups, projected calls/s/node, per call, separation vs tolerance, parity), basis: projected in prose (Q4 a; η = 1.0 declared, unmeasured until phase 4), the rule, secondaries, untested branches and caveats. The datacube/delaunay gpu_a100 hold is named, not recorded. Records are generated from the catalogue by the reviewed generator, never hand-typed.
Next (separate task, after autolens_profiling #496 merges): the producer side — build_catalogue reads the designations, sets class: optimal + decision_id on the cited rows, retires catalogue/provisional_designations.json.
Activity
Jammy2211 commentedon Oct 31, 2026
Heart RED override invoked by the human 2026-10-31 (AskUserQuestion: "Override + create issue"). Designations approved by the human 2026-10-31: "Approve all 39 (Recommended)"; datacube/delaunay gpu_a100: "Hold, keep provisional (Recommended)".
Merge pull request #84 from PyAutoLabs/feature/likelihood-lifecycle-p…
Jammy2211 commentedon Oct 31, 2026
Shipped — merged in #84 (
bcaf6b86, 2026-10-31, human merge command; Heart RED override recorded above).c218a428(now on main via #496).datacube/delaunaygpu_a100 held as provisional.pyauto-pulse checkOK.class: optimal+decision_id, retireprovisional_designations.json), then the day_zero stamp and snapshot.