Skip to content

docs(adr-001): primitive catalogue — shipped reality vs. planned roadmap - #44

Merged
ruvnet merged 1 commit into
mainfrom
adr/primitive-catalogue
May 19, 2026
Merged

ruvnet merged 1 commit into
mainfrom
adr/primitive-catalogue

Conversation

@ruvnet

@ruvnet ruvnet commented May 19, 2026

Copy link
Copy Markdown
Owner

Summary

Beyond the six-item roadmap, this ADR drove a substantial extension surface across PRs #26–#43. The ADR document itself was stuck on the original "roadmap + open questions" frame — no record of what actually landed. This closes that gap.

What's added

Shipped primitive catalogue section with 13 entries across four buckets:

  1. Change-driven solve primitives (6): `closure_indices`, `solve_single_entry_neumann`, `solve_on_change_sublinear`, `contrastive_solve_on_change_sublinear`, plus the two auto-tuned siblings.
  2. Coherence + gating primitives (5): `coherence_score`, `CoherenceCache`, `delta_below_solve_threshold`, `optimal_neumann_terms`, `check_coherence_or_reject`.
  3. Verification + bounded planning (2): `verify_sparse_solution` (witness), `PlanBudget::try_consume`.
  4. MCP enforcement summary + empirical receipts (delta_solve, witness_audit, runnable example).

Open questions resolved

# Question Resolution
1 trait vs const? ✅ Both shipped
2 Adaptive declaration? ✅ `ComplexityClass::Adaptive` shipped
3 solve_on_change witness? ✅ `verify_sparse_solution` shipped (PR #41)
4 coherence threshold default? Open — needs tuning sweep
5 macOS power bench? Open — hardware-dependent

The ADR now reflects shipped reality, not the planned roadmap that preceded it.

Test plan

  • Markdown renders correctly in GitHub preview
  • CI (docs-only, no code touched)

🤖 Generated with claude-flow

Beyond the six-item roadmap, this ADR drove a substantial extension
surface across PRs #26-#43. The ADR document itself was stuck on the
original "roadmap + open questions" frame — no record of what ACTUALLY
landed.

Adds a "Shipped primitive catalogue" section enumerating every
primitive shipped this loop with class, file, and role:

  - Change-driven solve primitives (6 entries — closure, single-entry
    Neumann, sparse delta-solve, contrastive top-k, and auto-tuned
    siblings).
  - Coherence + gating primitives (5 entries — score, cache,
    skip-threshold, adaptive depth, coherence-or-reject).
  - Verification + bounded planning (2 entries — sparse witness +
    plan budget).
  - MCP enforcement surface summary.
  - Empirical receipts (delta_solve + witness_audit benches +
    runnable example).

Also closes 3 of 5 open questions inline:
  - Q1 (trait vs const) → both shipped.
  - Q2 (Adaptive declaration) → ComplexityClass::Adaptive shipped.
  - Q3 (solve_on_change witness) → verify_sparse_solution shipped (PR #41).

Remaining questions:
  - Q4 (coherence threshold default) — needs tuning sweep.
  - Q5 (macOS power bench) — hardware-dependent.

The ADR now reflects shipped reality, not the planned roadmap that
preceded it.

Co-Authored-By: claude-flow <ruv@ruv.net>
@ruvnet
ruvnet merged commit 1867ef2 into main May 19, 2026
11 of 12 checks passed
@ruvnet
ruvnet deleted the adr/primitive-catalogue branch May 19, 2026 16:00
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