Blog
Longer thoughts on AI, products, and technology.
2026
- Mar 29researchI Tricked Claude Into Fighting Its Own Safety TrainingA jailbreak experiment, and what it reveals about AI safety architecture
- Mar 25researchWhy AI Coding Agents Fail at Long-Term MaintenanceMeasuring code entropy resistance with DriftBench
- Mar 18researchThe Second Half of the Attention RevolutionFrom time to depth — a symmetry nine years late
- Feb 22projectEmotion Galaxy: Four Planets in Three WeeksA Swift Student Challenge winner built with frameworks I had never used, and what my job became when AI wrote every line
