Writing

Notes on research, markets, and decisions, biased toward results that didn’t flatter me.

Why frontier models score <0.5% on ARC-AGI-3

2026-06-08 · 2 min read · draft

Notes from building a training-free system that matches frontier-model performance with zero LLM calls, and what the benchmark actually measures.

What a 1000× fee gap taught me about honest research

2026-06-01 · 5 min read · draft

I built a production-grade microstructure platform and the edge was real, just three orders of magnitude smaller than the fees. Why that's my favorite result.

Resign or play on: what chess gets right about quitting

2026-05-20 · 4 min read · draft

Chess players resign lost positions; researchers and traders mostly don't. On evaluation honesty, sunk costs, and knowing which kind of lost you are.