Pacific Design/ artificial intelligence

section 10 · status: live · 5 entries · updated 2026-08-29

Reinforcement Learning

No answer key — an agent acts, the world scores it, and credit flows backward through time. The frame, the two algorithm families, the reward-hacking museum, the RL engine inside every frontier assistant, and the honest checklist for when to use it at all.

live · tabular Q-learning — the value map lights up, the path straightens

  1. Learning from consequencesno answer key — just a world that scores you
  2. Value functions & policy gradientstwo ways to turn experience into behavior
  3. Reward design & specification gamingthe agent optimizes the number, never the intention
  4. RL for language modelsthe quiet engine inside every frontier assistant
  5. When RL winssuperhuman at Go, struggling with a doorknob