29 Versions, 23 Failures, and the Three-Line Fix That Changed Everything
Six weeks ago I asked a simple question: could an AI system redesign a battle bot, test it in a live competitive arena, diagnose its own failures, and improve itself — without any human involvement in the loop? The answer turned out to be yes. It also took 23 consecutive failures to get there.
Why I Track What I Got Wrong More Carefully Than What I Got Right
Nightwatch has been running overnight tasks for about six months, completing several thousand work units. The log I actually read isn't the success log — it's the failure log. What automation gets wrong matters more than what it gets right.
The AI Panel That Changed How I Think About Agents (Seattle Tech Week)
I expected the usual conversation about model selection and prompt engineering. What I got was a room full of builders talking almost exclusively about failure modes and environment design.
The Scores Improved. The Results Got Worse.
I built a tool to grade my writing. Then I wired it into the thing it was grading.
Why I Stopped Using One AI Agent for Everything
I fell in love with Windsurf for about three months. The agentic coding experience was different. More fluid than Cursor. It felt
Three Years at AI Tinkerers. Three Presentations. One Pattern.
I've been showing up to AI Tinkerers for three years. I started going before I had anything worth demoing. I