coulomb/frontend-patterns
5
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Audit prospective comparison readiness and retain effort gaps

Browse source 
Assistant: codex
Assistant-Model: gpt-6.1-sol
Assistant-Session: 01a1045b-933d-7aa1-b2db-104557487a79
This commit is contained in:
tegwick 2026-10-04 03:02:53 +02:00
parent 280c1d38bf
commit d8d3a22b54
9 changed files with 439 additions and 54 deletions
Download patch file Download diff file Expand all files Collapse all files

18
research/evolution/README.md
View file

@ -66,6 +66,9 @@ renew them. Historical snapshots retain their original schema-v1 assessments.
undermine their scope. Never promote automatically from metric thresholds.
Commands: `make evolution-report`, `make evolution-check`, `make evolution-test`.
Generation writes both EvolutionReport.md and the machine-readable
[ComparisonWindows.json](ComparisonWindows.json); checking rejects stale output
in either artifact.
The check rejects stale candidate hashes, missing references, inventory drift,
invalid provenance, unknown identities, corrupted baseline candidate hashes and stale
generated output. It does not
@ -112,6 +115,21 @@ baseline or the selected retrospective entries. Prefer more independent validati
when beliefs stabilize; broaden when new contexts expose missing decisions; repair
structure when ambiguity prevents use. Low novelty alone is not a stopping rule.
The window audit groups prospective entries by declared stream and comparison-group
label, orders them by date/identity, and retains links to every registered protocol.
Its three-cycle qualitative review trigger does not verify comparability: review
evidence tier, question granularity, evaluation and accounting against those protocols
first. Retrospective entries never join a window, even if later assigned effort or
a matching label. Outcomes include corroboration and inconclusive work.
The audit distinguishes positive research time, unknown time and recorded zero time.
Unknown or zero effort permits qualitative discussion with limits but prevents
effort comparison of the complete declared window. Do not silently discard those
cycles or reconstruct their time from commits. A narrower window needs an explicit
review explaining its boundary and retained exclusions. Readiness does not calculate
a trend, declare a plateau or complete WP-0006 T03; that task needs an evidence-bearing
review of useful comparable research, not a generated count.
Keep both the unchanged candidate cohort and additions/removals visible when
comparing snapshots. A growing inventory must not hide weakening evidence for old
claims. Learning continues after milestones; finite workplans carry the next bounded