Series
Structured learning paths and technical deep dives.
Read them in order, like chapters in a book.
Building an Agent
A comprehensive series on building AI coding agents from scratch, covering everything from basic tool integration to advanced features like agentic search and subagents.
Building An Agent
Implementing a coding agent in around 200 lines of Javascript code
Building a Coding CLI with React Ink
Migrating Our Coding Agent to React Ink
Building your own AI SDK
Switch between models with your own custom router
Agentic Search
Models can't do much without the right context, agentic search does just that
LLMs From Scratch
Working through Stanford's CS336 and documenting what I learn at each step — from tokenization to training.
Openclawd From Scratch
Build Your Own OpenClaw From Scratch is a hands-on series where we rebuild an OpenClaw-style agent from first principles. Starting from a blank file, we'll implement the core agent loop, tool use, sandboxed execution on Modal. By the end, you'll have an agent that can install what it needs, run real workflows, and evolve over time.
The Lay of the Land
Breaking down OpenClaw-style agents and the roadmap for building one from scratch
It's alive!
Allowing an agent to modify itself with tools
ET Phone Home
Deploying our agent on Telegram with ngrok so it can report back
Total Recall
Persisting conversations to a database and compacting context so our agent never forgets
Deep Research From Scratch
A series on building deep research workflows from scratch, covering the core components needed to plan, search, synthesize, and present findings.
Reinforcement Learning From Scratch
A series on building reinforcement learning agents from scratch, covering everything from basic concepts to advanced techniques.
Structured Outputs From Scratch
Building a structured output parser from first principles — from JSON Schema to regex to FSM to constrained decoding.
Pydantic to Regex: Compiling JSON Schemas for Structured Outputs
Build a recursive JSON Schema-to-regex compiler from first principles
Building our FSM
Going beyond a simple regex to an FSM for quick lookups
Compiling IR to NFA
Turn regex IR into a graph-based automaton for fast next-step lookup