June 2, 2026 · 2:07 PM 9 min read
The model writes faster than I can read, so I stopped reading
I built a self-healing agent loop that writes code, finds bugs, fixes them, reviews fixes, and merges—all without a human in the cycle. Here's how I actually kept it working: I reviewed tests, not code, because that's where my judgment still beats the model.
May 12, 2026 · 10:12 AM 5 min read
Fluency is a reflex
I help teams build AI fluency—not through training programs, but by showing them how to think differently about problems they actually care about. What I learned is this: the reflex develops in your personal life first, then shows up at work automatically.
May 4, 2026 · 8:24 AM 6 min read
What goes wrong when you tell AI to stop being nice
I spent three months using a provocative personalization prompt designed to make Claude push back on me constantly. Here's what actually changed—and where the technique backfires.