Reading up on AlphaGo

Coding agents combined with reasoning LLMs have become automated scientists, enabling a golden age where all computer science problems appear tractable through vast inference compute.

Building AI agents that replicate human workflows traps them in a local optimum; true efficiency requires designing around the problem's objective function instead.

Zhengdong Wang argues that compute scaling—underpinned by Moore's Law and empirical trends—has driven AI progress more reliably than human innovation, despite repeated underestimation.

Reinforcement learning provides far fewer bits per FLOP than pretraining until models achieve high pass rates, limiting RLVR's ability to learn new capabilities.