docs: master narrative (single source of truth) + retire stale 2361 test count#12
Draft
jessekemp1 wants to merge 2 commits into
Draft
docs: master narrative (single source of truth) + retire stale 2361 test count#12jessekemp1 wants to merge 2 commits into
jessekemp1 wants to merge 2 commits into
Conversation
…61 test count NARRATIVE.md is the canonical description of Cortex: frozen one-liner, claims table (the only numbers any artifact may quote), capability map, story beats, honest comparisons, weaknesses, asks, and a derivation map for README / Show HN / talks / beta onboarding. Also corrects the two remaining live references to the retracted 2,361 test figure (README Testing section, Show HN draft) to the audited 1,855 fresh-clone number. Historical mentions in CHANGELOG and AUDIT_FINDINGS are intentionally preserved.
Adversarial code-verification pass over the presentation material found several claims that outran the implementation. Corrections: - Anti-patterns: what's real is 37 seed anti-patterns surfacing as pre-task warnings (70% of injected-context budget); there is no type-based retrieval boost, and custom gotcha capture is write-only today. README/NARRATIVE now say exactly that. - Memory tiers renamed to match code (short-term -> working -> long-term; 'episodic/semantic' appear nowhere in tiered_memory.py) and promotion into the permanent tier is documented as explicit indexing, not automatic (the promotion rule is currently a logging no-op). - 'Nothing leaves your machine' scoped honestly: the store is local and telemetry-free; LLM-backed queries call Anthropic, Voyage/notifications are outbound only if configured. - Removed the '<800 tokens' routing threshold (no such threshold exists; the classifier scores keywords/scope/char-length) and the unbenchmarked <10ms/<100ms/<1s latency ladder (LLM-backed queries measure ~10s). - Data-storage tree replaced with the layout the code actually writes. - cortex demo no longer references com.cortex.outcome-linker (no such agent ships); it points to the manual linker invocation. New: docs/USING_CORTEX.md — operator's guide grounded in file:line verification: install paths (incl. the bare 'pip install -e .' trap that omits the bridge/MCP), the daily loop, config surface, and the compounding levers (BM25 indexes only decision tags, which the MCP tool cannot set — write searchable prose; always pass project=; GOALS.md grammar per supervisor/intake.py; reindex staleness; honest what-doesn't-work-yet list).
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
docs/NARRATIVE.md— the canonical master narrative every artifact derives from: frozen one-liner + elevator, problem statement, a claims table that is the only place numbers may be quoted from (each with source, verified date, and caveat), the four-primitive capability map, chronological story beats (origin → depth-over-speed → strategic pivot → the audit → productization), honest Mem0/Supermemory comparisons, unflinching weaknesses + disruption risks, asks by audience, and a derivation map (README ← §1/2/4, Show HN ← §1/2/6/8, talks ← all, etc.).2,361test count (README.mdTesting section,docs/show_hn_draft.md) to the audited 1,855 fresh-clone number. Historical mentions inCHANGELOG.mdanddocs/AUDIT_FINDINGS.mdare preserved on purpose — they document the correction itself.Why
Foundation for the upcoming peer presentation and all future Cortex storytelling (README iterations, Show HN, beta onboarding). The claims-table rule exists specifically to prevent another 2,361-style drift: update the table first, then sweep derived artifacts (
grep -rn "<old number>" README.md docs/).This pull request and its description were written by Isaac.