Learning as Approximation
A learning note on the shared correction pattern behind gradient descent, temporal difference learning, stochastic approximation, and Bellman fixed-point methods.
Longer-form essays and reflections: where ideas have settled enough to be argued.
A learning note on the shared correction pattern behind gradient descent, temporal difference learning, stochastic approximation, and Bellman fixed-point methods.
A concrete design note on my personal AI workbench: owners, state flows, cron maintenance, instruction boundaries, and a Karpathy-style Obsidian second brain adapted for daily work.
A short learning note on how word representations led to attention, self-attention, and Transformers.
A practical system note on routing scattered LLM API keys through a self-hosted New API gateway, then exposing that gateway across devices with Tailscale.
A historical note on what my OpenCode setup taught me about agent roles, permission boundaries, protocol interfaces, and cross-platform configuration.
A long-form learning note on data abstraction, from Java interfaces and C++ concepts to TypeScript unions, Kotlin sealed hierarchies, and Haskell algebraic data types.
A historical note on how I used Nix, Home Manager, and modular configuration to keep a NixOS desktop and macOS laptop aligned.
A short note on why I keep a public research notebook instead of treating writing, notes, and project traces as separate scraps.