Writing
DraftRead the draft →
Deriving Reinforcement Learning from the Bellman Equation
Bellman as a fixed point, the dual linear program and the state–action polytope, and where online, offline and language-model RL sit inside that picture.
DraftRead the draft →
Deriving the Prior-Data Fitted Network
From Bayesian prediction to tabular in-context learning, Do-PFN, and time-series forecasting. With figures you can move.
Notes
DraftRead the note →
Reinforcement learning: a synthesis
A long working note from MDPs and the Bellman equation to policy gradients, offline RL and GRPO, with every claim marked derived, cited or recalled.