The auditability library.
Field notes on proving what AI agents did — hash-chained logging, incident evidence, compliance logging, and multi-agent causality. Written from building the recorder, not from press releases.
Hash chains, append-only ledgers, and one-click verification — the mechanism that turns agent logs into evidence.
2026-10-11 · 7 min read
The 2 AM playbook: freeze evidence, reconstruct the cited timeline, and answer the five questions every stakeholder asks.
2026-10-11 · 8 min read
Observability answers 'what is my agent doing?' Tamper-evident logging answers 'can you prove what it did?' — different jobs, different instruments.
2026-10-11 · 6 min read
Automatic logging, full operation coverage, tamper protection, independent verification, retention — mapped to real ledger mechanics.
2026-10-11 · 7 min read
Timing tells you what happened near something. Declared causal links tell you what happened because of something.
2026-10-11 · 6 min read