1 nota · última: 23 Sep 2026
Argues for building toy programs (regex engines, OS kernels, chess AIs) as the fastest route to real understanding, and against outsourcing that struggle to an LLM.