Blog
Articles on AI agent debugging, testing, and observability.
I Turned Jensen Huang's Physical AI Framing Into a Claude Code Skill
Most product plans fail for the same reason most AIs fail: they operate in a frictionless vacuum. Jensen Way is an AI coding skill that forces the causal chain, scores six laws of reality, and produces a Build / Pivot / Kill verdict backed by evidence. Works with Claude Code, Cursor, Codex, Windsurf, and any assistant that reads context files.
Your agent failed at step 30. rewind fix diagnosed and repaired it in 12 seconds.
Introducing rewind fix — AI-powered diagnosis and repair for failed agent sessions. One command finds the root cause, suggests a fix, and verifies it works.
Introducing Rewind: Chrome DevTools for AI Agents
AI agents ship with 30-step chains, but debugging them means re-running everything from scratch. Rewind records every LLM call, lets you fork at any failure, replay with the fix, and prove it works.
Debugging Langfuse traces with Rewind
Import a Langfuse trace by ID, fork at the failure, replay with the fix, score with LLM-as-judge, and export back. A step-by-step integration guide.