Skip to content

Park Phase 6 in the roadmap, and put dogfooding first#326

Open
samrusani wants to merge 1 commit into
mainfrom
docs/park-phase6
Open

Park Phase 6 in the roadmap, and put dogfooding first#326
samrusani wants to merge 1 commit into
mainfrom
docs/park-phase6

Conversation

@samrusani

Copy link
Copy Markdown
Owner

The roadmap claimed Phase 6 work had not begun. It was attempted across three reviewed build-and-review increments and is preserved unmerged on codex/phase6-counting-substrate.

Records honestly:

  • What the substrate achieved: write-time occurrence identity, provable sums with provenance, review gating, no reachable exact answer, extraction from ordinary English moved off near-zero, query side parses or refuses cleanly, relative date windows resolve. Adversarial review found and removed a mechanism that would have reported confidently wrong counts.
  • Why it is parked: the deterministic extractor recovers too few events from real prose to reach the target, and the query and write stages are each short of it independently. The remaining gap is semantic event extraction from arbitrary English.
  • The unused lever: model assistance at capture time under the existing review-gated proposal pattern, always permitted by the design and never attempted. The rule against model calls in pack compilation is unaffected.

Dogfooding moves to first position because counting is currently justified by a benchmark rather than observed use.

Docs only. Versions untouched. Control-doc truth and its test suite pass locally.

The roadmap said Phase 6 work had not begun. It was attempted across three
reviewed increments and is preserved unmerged on its branch. Record what the
substrate achieved, why it is parked rather than abandoned, and the one lever
the design always permitted but no increment used: model assistance at
capture time under the review-gated proposal pattern, with the rule against
model calls in pack compilation unaffected.

Dogfooding moves ahead of it, because counting is currently justified by a
benchmark rather than by any observed use, and a live instance with no
memories cannot say whether counting is the capability that matters most.

@greptile-apps greptile-apps Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Your trial has ended. Reactivate Greptile to resume code reviews.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant