docs(opmem): publication readiness package (13-task pins, no re-run)

[sdntsng]

Summary

OpMem-first publication package built only from existing repo evidence. No benchmarks were run and no provider spend. The goal is to state clearly what is ready for a Paper 1 submission versus what is still missing, and to align normative docs with the live 13-task suite (upd03).

Independent static review confirms the frozen 2026-08-15 pin: Brainy 13/13 and Mem0 Platform 10/13 are supported by markdown artifacts and Brainy per-task JSON; we do not claim re-validation on current bbe55f7 or inflate Platform numbers as OSS-reproducible.

Changes

• docs/research/opmem/PUBLICATION_READINESS.md — full package:
  • Reproducible protocol (operations, adapters, commands, pin fields)
  • Exact claims / evidence table (allowed vs forbidden wording)
  • Independent 13/13 vs 10/13 evidence review (fixture count, artifacts, CI caveat)
  • Missing experiments and ablations
  • Threats to validity
  • Artifact instructions for reviewers
  • Manuscript outline
  • Overlap map with draft PRs docs: n=1540 product-recall handoff for next-cycle planning #185–S5: versioned RRF retrieval for search and /recall #187 (and pointer to inventory docs: inventory papers, protocol pins, and reproduction plan #188)
• docs/research/opmem/README.md — index and pin summary
• Stale 12→13 task wording (provable only):
  • docs/research/opmem-spec.md — taxonomy includes upd03, 13 tasks
  • docs/research/public-bench-ladder.md — L0 OpMem row → 13/13 vs 10/13
  • docs/research/mem0-parity-gaps.md — executive OpMem row
  • docs/research/master-plan.md — §1.1 OpMem row
  • docs/research/posts/2026-07-opmem-v0.md — supersession banner (July 12/12 numbers preserved as historical)

Intentionally not rewritten: dated artifacts (opmem-baseline-report.md, opmem-staging-vs-mem0.json, launch-narrative.md, etc.) remain historical 12-task records.

No product code changes.

Test plan

• find fixtures/opmem -name '*.json' | wc -l → 13
• opmem-spec.md no longer says v0: 12 tasks
• go test ./... — not required (docs-only)
• OpMem re-run on bbe55f7 — explicitly out of scope (no spend)

Linear / tracking

OpMem Paper 1 track (paper-topics.md). Inventory cross-ref: PR #188 (inventory-reproduction-plan-2026-09-18.md).

Risk / follow-ups

• Not publication-ready: no LaTeX manuscript, no ≥4-system table, no committed 3-system JSON for the 2026-08-15 pin, no bbe55f7 re-measure.
• CI does not assert 13/13 — run_opmem.py exits 0 on task failures; documented as a threat.
• PR S5: versioned RRF retrieval for search and /recall #187 (RRF): merge should trigger OpMem re-run before claiming ops non-regression.
• Do not merge to main without explicit approval (staging = dev only).

Ready vs missing (from package)

Ready	Missing
13 fixtures + runner + 3 adapters	Zep/Letta/LangMem; v1 ~30 tasks
Protocol + spec + CI harness health	OpMem RunManifest / frozen multi-system JSON in git
13/13 vs 10/13 markdown + Brainy integrity JSON	Mem0 per-task JSON in git; OSS run; SHA bbe55f7 re-pin
Manuscript outline + threats	Camera-ready paper, related work, external fixture audit