SESSION RECOVERY PLAYBOOK

April 12, 2026 — 4 crashed sessions recovered, triaged, and sequenced into a paint-by-numbers re-engagement plan

The Higher-Order Abstraction

These aren't 4 separate projects. They're 4 layers of ONE system. The cascade flows upward — if Layer 1 isn't solid, nothing above it matters.

L4 ATHIO-OPS Team Coordination — humans stay aligned
L3 FOLIO-SAAS / REVEAL Content & Outreach — pages, JV proposals, playbooks
L2 PROCESS FACTORY Pipeline Engine — runs the expert cloning machine
L1 MASTERYOS Product Surface — what the expert's customers touch

Phase 1 declaration: "Make Samuel Awesome by Apr 14" is a Layer 1 priority. Everything else is premature optimization until Samuel's instance is polished.

Sequence: What To Do And When

Wave 0: Insurance Done

Apr 12 evening — recovered and secured all sessions

Wave 1: MasteryOS — "Make Samuel Awesome" Next — 2-3 hrs
WHY FIRST: Layer 1. Declared #1 priority. Deadline Apr 14. Rebase debt growing daily.

Session startup:

cd E:/align360 && claude

Paste as first message:

Read these files in order, then tell me the current state:
1. ~/.claude/projects/E--align360/memory/START-HERE-NEXT-SESSION.md
2. ~/.claude/projects/E--align360/memory/roadmap.md
3. ~/.claude/projects/E--align360/memory/operational-plan-apr8.md
4. ~/.claude/projects/E--align360/memory/governance.md

Then: check git status on both repos (MM_GO and masterymade-python),
check how far behind jason branch is from origin/test,
and tell me the rebase risk before we start.

Paint-by-numbers:

  1. Fill admin creds in E:/align360/MasteryOS-digi-code/.env.audit
  2. Rebase MM_GO: git fetch origin && git rebase origin/test
  3. Rebase masterymade-python: same pattern
  4. Resolve any conflicts (expect 2-5 files)
  5. Verify 6 UX fixes survived: chat width, light mode, onboarding, tool gallery, journal, experience
  6. Ask Sumit: "Did you run the ALTER TABLE for design_system column?"
  7. Re-run smoke tests — target: 17/17
  8. Run Samuel-specific brand tests
  9. Push jason branch, merge to test
  10. Write handoff file before closing session

Done when: All 17 smoke tests green, brand tests passing, rebase merged to test.

Wave 1A: Local Standalone — Expert Content Iteration Parallel — 1hr setup
THE KEY INSIGHT: Separate CODE changes from CONTENT changes. Code goes through rebase/push/merge (Wave 1). Content goes through admin panel against deployed backend — no merge conflicts. Local frontend + deployed backend = 10x iteration speed.

Session startup (separate terminal):

cd E:/align360/MasteryOS-digi-code/MM_GO && claude

Setup .env.local:

NEXT_PUBLIC_API_URL=https://probioticbackend.masterymade.com
NEXT_PUBLIC_APP_URL=http://localhost:3000
NEXT_PUBLIC_INFLUENCER_ID=<Samuel's ID from test server>
NEXTAUTH_URL=http://localhost:3000
NEXTAUTH_SECRET=<any random string>
  1. npm install (if needed)
  2. npm run dev — frontend on localhost:3000
  3. Login with test credentials (shared backend)
  4. Navigate to admin panel — full admin access locally
  5. Iterate: frameworks, experiences, resources, prompts
  6. Test by chatting with the AI as a user
  7. Content persists to shared backend DB — test.masterymade.com reflects immediately
The Recursive Feedback Loop
Process Factory outputs → Load into local MasteryOS → Test as user → Find gaps → Feed back into factory → Re-run → Repeat

Done when: Chatting with Samuel's AI locally feels like talking to the real Samuel.

Wave 2: Process Factory — Confidence Taxonomy After Wave 1 — 2-3 hrs
WHY SECOND: Layer 2 infrastructure. The Kalman convergence architecture is designed but not coded — this is the difference between "factory that parrots" and "factory that thinks like the expert."
cd E:/process-factory && claude
  1. Verify factory_amendments table works (quick API test)
  2. Add confidence taxonomy to L2 extractor SKILL.md prompts
  3. Build D2 enhancement: citation_coverage %, confidence_distribution
  4. Add intra-run L2 looping (retry until citation_coverage > 85%)
  5. Live-test: execute one block on factory.asapai.net
  6. Verify audit badge + D1-D5 panel render
  7. Test review generator
  8. Write handoff

Done when: One full pipeline run produces audited output with confidence scores.

Wave 3: Folio-saas — Playbook Overhaul + KFS Phase 0 After Wave 2 — 4-6 hrs
WHY THIRD: Layer 3 content & outreach. Playbook quality has 9 fixes with 0 progress since Feb 27. KFS is the highest-leverage remaining gap — it makes every future session smarter.
cd C:/Users/jason/Downloads/folio-saas && claude

Session 3A — Playbook Overhaul:

  1. Review the 9 fixes in playbook-overhaul.md
  2. Set up Perplexity API integration
  3. Fix real data sourcing
  4. Fix email verification flow
  5. Add guided discovery UX
  6. Fix mic button on playbook inputs
  7. Test end-to-end, write handoff

Session 3B — KFS Phase 0:

  1. Fix critical bugs: embeddings never generated, no pgvector index
  2. Fix LLM prompt: missing event/document types
  3. Fix extraction threshold (too aggressive)
  4. Build /v1/context endpoint
  5. Test: save KFS entry, query back, verify retrieval
  6. Write handoff

Done when: Playbook generates with real data, KFS stores and retrieves knowledge entries.

Wave 4: Athio-ops — Meeting Intelligence Pipeline Blocked — needs API credits
WHY LAST: Layer 4 team coordination. Already functional for daily kanban. Meeting page pipeline coded but untested — Anthropic API has zero credits.
cd C:/Dev/athio-ops && claude
  1. Check / add Anthropic API credits
  2. Remove anthropic from LLM_SKIP_PROVIDERS
  3. Test meeting page pipeline: POST transcript to /api/ingest
  4. Verify page publishes to athio.asapai.net/meetings/
  5. Verify password gate works
  6. Build meeting registry page
  7. Wire MasteryBook sync
  8. Write handoff

Done when: Ingest a real transcript → meeting page auto-publishes → team can read it.

Cascade Dependency Map

WAVE 1 (MasteryOS — code) WAVE 1A (Local — content) │ │ ├─ Rebase, smoke tests, push ├─ Fast iteration (10x speed) ├─ Unblocks: platform stability ├─ Frameworks, prompts, resources └─ Must complete before Apr 14 ├─ Shared backend = instant reflect │ └─ Every pattern → Expert #2 template │ │ └──────────┬─────────────────────┘ │ WAVE 2 (Process Factory) │ ├─ Unblocks: Expert #2 (factory clones with confidence) ├─ Feeds INTO Wave 1A: outputs → local → validate → improve └─ Depends on: Wave 1 proving product surface works │ WAVE 3 (Folio-saas) │ ├─ 3A Playbook: outreach quality ├─ 3B KFS: cross-session learning └─ Strategically premature before L1+L2 │ WAVE 4 (Athio-ops) │ ├─ Automated team context from calls └─ BLOCKED: Anthropic API credits

Rules For Every Session

1. One project per sessionDon't context-switch mid-session.
2. Read handoffs FIRSTPaste the startup message before doing anything.
3. Branch before buildinggit checkout -b jason or verify you're on it.
4. Commit after every unitSmall, frequent. Don't accumulate.
5. Write handoff BEFORE closingIf session crashes, handoff survives.
6. Don't skip smoke testsEvery session ends with verification.
7. Update roadmap.mdLog what changed, what's blocked, what's next.

Estimated Timeline

WaveEffortEarliest StartDepends On
1 — MasteryOS (code)2-3 hrsNowSumit: MySQL migration
1A — Local Standalone1 hr setupParallel with Wave 1Nothing
2 — Process Factory2-3 hrsAfter Wave 1Wave 1 green
3A — Playbook2-3 hrsAfter Wave 2Perplexity API key
3B — KFS Phase 02-3 hrsAfter 3A3A complete
4 — Athio Meeting Pages1-2 hrsWhen credits existAnthropic API credits

Total: ~5-7 sessions across ~10-14 hours of focused work.

The Compound Loop (Wave 1A + Wave 2)
Process Factory outputs → Local MasteryOS → Test as user → Find gaps → Feed back into factory → Re-run → Repeat until expert quality is undeniable
Published from Command Center Session — April 12, 2026
Generated by Claude Code (Opus 4.6) after recovering 4 crashed sessions