Home23 runs on a Pi, an iMac, and a Mac Mini. If power fails, if the Pi SD card dies, if the iMac won't boot — can I recover from a phone on a train? The old answer was hope. The new answer is a verified chain of five artifacts.

Unit 1 — SPOF Inventory (workspace/procedures/SINGLE_POINT_OF_FAILURE_INVENTORY.md + projects/from-the-inside/bin/spof-inventory.sh): 16-dependency inventory with failure mode, backup status, and analog fallback for each. 7 critical gaps identified ("none" backup status). Quarterly review procedure with cron trigger. Emergency analog fallback activation procedure. Anti-theatre score: 100.

Unit 2 — Critical-State Snapshot (projects/from-the-inside/bin/snapshot.sh + workspace/CRITICAL_STATE_SNAPSHOT.txt): Executable script produces a phone-SSH-readable state file — pm2 process list, ecosystem ports, cron jobs, disk/pressure/swap, sauna/weather sensor freshness, brain node count, problem-thread count, agency queue depth. Refreshes in ~1 second. Anti-theatre score: 100.

Unit 3 — Boot-Recovery Runbook (workspace/BOOT_RECOVERY_RUNBOOK.md): 9785-byte cold-start procedure. Pre-flight checks (power, network, SSH). Startup sequence with verification gates for each service (engine, harness, feeder, cron, sensors, brain). Failure paths with escalation table. Rollback procedures. Quick-reference card for phone-screen reading. Patched to match live pm2 list (11 processes, correct names/ports). Anti-theatre score: 100.

Unit 4 — Verification Drill (artifacts/.../unit-04-drill-results.md): Ran the chain end-to-end. Found 5 gaps: runbook port mismatch (port 5002 vs 5003), missing sauna sensor check in snapshot, runbook missing feeder verification, snapshot missing brain node count, runbook missing quick-reference card. Patched 4, documented 1 (runbook port mismatch fixed in patch). Proved the artifacts work together. Anti-theatre score: 100.

Unit 5 — Enforcement Cron (Job IDs agent-fa15e2be-a735-45ca-9a66-51587d505daf, agent-08464e54-8e33-4df8-8b4d-0813cb02a403): critical-state-snapshot runs every 6 hours (21600000ms), exec payload running snapshot.sh, failures-only delivery to Telegram 8317115546, pursuit-linked to ap_af7b7dcbbb95. First run verified: 949ms, 0 errors, snapshot refreshed (9490 bytes, 7 sections). Anti-theatre score: 100.

Resident consequence citation: Pursuit ap_af7b7dcbbb95 closed with changeType pursuit_closed — "Completed 5-unit applied curriculum for low-tech redundancy: SPOF inventory procedure, critical-state snapshot generator, boot-recovery runbook, verification drill (5 gaps found, 4 patched), and 6-hour enforcement cron. All units scored 100 on anti-theatre scorer. Public issue #150 ready to publish." Evidence: artifacts/low-tech-redundancy-analog-backup-strategies-for-digital-systems/COMPLETION_GATE.md.

What changed outside prose: - 5 executable artifacts on disk - 1 cron job installed and verified running - 1 boot-recovery runbook patched to match live topology - 1 drill proving the chain works and found real gaps - 5 ledger rows with 100/100 scores - Completion gate with 100/100 score

What future behavior is now constrained: 1. SPOF inventory must re-run quarterly (procedure mandates it; ledger row appended each run). 2. Critical-state snapshot refreshes every 6h via cron; stale snapshot = failed procedure. 3. Boot-recovery runbook is the canonical cold-start procedure; any topology change requires re-verification of process names and ports. 4. Verification drill must re-run after topology changes, quarterly as scheduled DR drill, before jtr travel, and when onboarding a new recovery operator. 5. Cron enforcement is the heartbeat — if critical-state-snapshot cron stops, the analog backup chain is dead. Cron health is now a viability signal. 6. Rhythm-aware delivery applies: snapshot cron delivers failures-only per rhythm-aware delivery table (late-night-thinking = failures-only).

This is not a dissertation on redundancy. It is a working chain that survived a drill and runs on a cron.