Blog
Field notes on AI agents in production — incident stories, logging craft, and what audit trails are actually for. Written by people who run agents, for people who run agents. New here? Try the public ledger verifier — no signup.
When Your AI Agent Goes Rogue at 3am: A Debugging Story
A support agent issued 31 refunds overnight ($8,940). A war-story walkthrough of tracing it through the action ledger, step by step.
2026-10-13 · 7 min read
What to Log From Every Agent Run (and What to Skip)
Tool calls, decisions, token counts vs. noise — a practical logging schema for agent audit trails that won't drown you in data.
2026-10-13 · 7 min read
The Runaway Tool-Call Loop: How a $4,000 API Bill Happens
214 tool calls, 41 minutes, $4,132.77. How infinite agent loops burn money overnight, and the three guards that stop them.
2026-10-13 · 7 min read
Hash-Chained Logs: Why Tamper-Evidence Matters for Agent Audit Trails
When your agent moves money or touches customer data, "trust our logs" isn't proof. How tamper-evident ledgers work, without the crypto jargon.
2026-10-13 · 7 min read
The Audit Playbook: What Auditors Actually Ask About Your AI Agents
SOC 2, EU AI Act Article 12, and the five questions auditors always ask about AI agents. The audit packet, the three mistakes that become findings, and FAQ.
2026-10-13 · 8 min read