Sinew Benchmark Experiments — 6 experiments to produce irrefutable capability proof. Start with pure-function experiments (anti-jailbreak, trust dynamics, crisis detection). CRITICAL — determines valuation and strategy. See [projects/sinew-benchmark-experiments.md](projects/sinew-benchmark-experiments.md)
Sinew IP Protection — Open core transition: split open primitives from proprietary Person model, trust, rate limiter. License decision + repo restructure + landing page update. CRITICAL — before any public launch. See [projects/sinew-ip-protection.md](projects/sinew-ip-protection.md)
Kin Onboarding — Invisible three-phase design: earn second conversation, accumulate toward aha moment, reveal Person model. See [projects/kin-onboarding.md](projects/kin-onboarding.md)
Sinew Web Separation — Split sinew.dev into framework site (no auth) + Kin product app (separate domain). Blocks launch credibility. See [projects/sinew-web-separation.md](projects/sinew-web-separation.md)
Context Refresh Architectures — Research Paper. Formalize CRA framework, design experiments, collect data, publish. Blocked: OS system must reach steady state first. See [projects/context-refresh-paper.md](projects/context-refresh-paper.md)
activeMaster Launch Plan
activevoice calibration
activedesign system bootstrap
activeFinancial Independence via Talkable Apps
activekin milestone v1
activekin onboarding
activelaunch readiness
activenerv mvp
activetier aware routing enforcement
activesystem consistency metric
activesinew web separation
activesinew valuation analysis
activesinew person kin architecture
activesinew ip protection
activesinew experiment results
activesinew benchmark experiments
activeproduct vision deep research
activeproduct vision analysis
activePerson Model — Cognitive TDD Implementation
activeos infrastructure for kin
activeNous — System Architecture Plan
activeNous — North Star
activenerv ux build
activecognitive enforcement layer
seedCognitive Infrastructure Ecosystem
activecognitive runtime strategy
activeConductor Infrastructure Plan
activecontext refresh paper
activeconversation memory architecture
activedashboard in sinew analysis
activedashboard in sinew research
Semantic routing NaN scores — root cause: 1007 OpenAI embeddings (1536-dim) mixed with Gemini (3072-dim). Fix: model filter at query time + re-embedded all with Gemini. 7435/7435 now uniform.16 Mar
Embedding provider fallback — investigate Grok as alternative/fallback for Gemini embeddings. Result: Grok has no standalone embeddings endpoint (locked behind managed Collections API). Not usable. Alternatives if needed: OpenAI text-embedding-3-large, Cohere embed-v4.16 Mar
Multi-key API rotation for embeddings — 3 Google accounts configured, auto-discovered from GOOGLE_API_KEY_N_label pattern, round-robin + rotate on 429.16 Mar
Fix phone Obsidian — clone `os-notes` repo on phone, point Obsidian vault to it. See [plans/phone-obsidian-notes-repo.md](../plans/phone-obsidian-notes-repo.md)15 Mar
Populate LW Explorer semantic embeddings — run `cd dashboard && npx tsx scripts/sync-lw-embeddings.ts` when Gemini rate limits allow. 3287 entities, saves progress in batches of 50. Can be re-run safely (skips unchanged).15 Mar
New Laptop Setup
Sinew — MVP Product Spec
Sinew — Monetization Strategy
Foresight
reminderContinue virtual company build: Ops Monitor is deployed and filed first report (3 products unreachable). Next steps: (1) Wire cron trigger for 30-min auto-runs, (2) Build Team Activity dashboard page, (3) Build second employee (Research Agent). Plan: context/plans/virtual-company-architecture.md. Aliases ready: ops-monitor, consolidate, quorum-check, decay, synthesis.→
reminderKin has no DB migration system — schema.sql changes don't propagate to running DBs (local or production). The role column incident proved this. Build either: (1) a migrations/ directory with numbered SQL files and a startup runner, or (2) a startup diff that compares schema.sql column definitions against information_schema and warns/auto-applies. Repo: ~/dev/kin, schema: src/server/schema.sql, DB init: src/server/db.ts→
reminderReview post-compaction recovery effectiveness data. V7 deployed 2026-03-12 — check Recovery Effectiveness card on /reliability for production metrics. Compare production signal/notes/learnings rates against benchmark predictions (85%/100%/100%). If production diverges significantly from benchmark, investigate whether the full context (CLAUDE.md + compaction summary + recovery text) creates different dynamics than the isolated benchmark. Update cognitive-principles.md #12 and #13 with production findings.→
taskContext Refresh Architectures — Research Paper. Formalize CRA framework, design experiments, collect data, publish. Blocked: OS system must reach steady state first. See [projects/context-refresh-paper.md](projects/context-refresh-paper.md) — active for 22 days→
connection6 signals about publication in 30 days but no file updates — knowledge may not be captured→
connection39 signals about core in 30 days but no file updates — knowledge may not be captured→
connection7 signals about context/chief-of-staff/voice-interface.md in 30 days but no file updates — knowledge may not be captured→
connection5 signals about kin/product in 30 days but no file updates — knowledge may not be captured→
connection8 signals about ops/docker in 30 days but no file updates — knowledge may not be captured→
connection25 signals about context/publication in 30 days but no file updates — knowledge may not be captured→
connection15 signals about context/plans in 30 days but no file updates — knowledge may not be captured→
connection7 signals about context/tasks/projects/cognitive-infrastructure-ecosystem.md in 30 days but no file updates — knowledge may not be captured→
connection9 signals about core/process in 30 days but no file updates — knowledge may not be captured→
connection7 signals about context/initiatives in 30 days but no file updates — knowledge may not be captured→
connection6 signals about research/social-simulation in 30 days but no file updates — knowledge may not be captured→
connection6 signals about publication/experiment-log in 30 days but no file updates — knowledge may not be captured→
connection16 signals about sinew/product-strategy in 30 days but no file updates — knowledge may not be captured→
connection5 signals about publication/substack in 30 days but no file updates — knowledge may not be captured→
connection9 signals about career/learnworlds in 30 days but no file updates — knowledge may not be captured→
connection19 signals about context/research in 30 days but no file updates — knowledge may not be captured→
connection8 signals about context/sinew in 30 days but no file updates — knowledge may not be captured→
connection8 signals about soma in 30 days but no file updates — knowledge may not be captured→
connection18 signals about sinew in 30 days but no file updates — knowledge may not be captured→
connection11 signals about publication/threads in 30 days but no file updates — knowledge may not be captured→
connection8 signals about context/ops/security-audit.md in 30 days but no file updates — knowledge may not be captured→
connection11 signals about context/dev-skills in 30 days but no file updates — knowledge may not be captured→
connection10 signals about ops/infrastructure in 30 days but no file updates — knowledge may not be captured→
connection7 signals about sinew/nerv in 30 days but no file updates — knowledge may not be captured→
connection7 signals about people in 30 days but no file updates — knowledge may not be captured→
connection6 signals about career/learnworlds/recommendations-kb in 30 days but no file updates — knowledge may not be captured→
connection12 signals about context/publication/threads/compaction-survival.md in 30 days but no file updates — knowledge may not be captured→
connection6 signals about ops/deploy in 30 days but no file updates — knowledge may not be captured→
connection5 signals about engineering in 30 days but no file updates — knowledge may not be captured→
connection12 signals about context/core in 30 days but no file updates — knowledge may not be captured→
connection28 signals about ops in 30 days but no file updates — knowledge may not be captured→
connection12 signals about publication/research in 30 days but no file updates — knowledge may not be captured→
connection34 signals about context/ops in 30 days but no file updates — knowledge may not be captured→