Brain Drip

An open notebook on modern machine learning.

Four kinds of writing live here. Drips are interactive explainers for single concepts. Courses are sequences of Drips, read in order. Notes are essays and observations. Blueprints are practical guides with working code. Everything is free.

Recent entries

  1. May 20, 2026NoteAI Agents Work. The Data Doesn't.
  2. May 14, 2026DripNatural Language Autoencoders9 min
  3. May 14, 2026NoteState of OSS AI, Spring 2026
  4. May 13, 2026NoteThe /goal command — 11 days, 3 tools, one primitive
  5. May 9, 2026NoteWhy I'm redesigning Brain Drip6 min
  6. May 9, 2026NoteAirbnb's 60% playbook
  7. Apr 22, 2026DripTransformers, From First Principles14 min
  8. Apr 14, 2026DripDiffusion Models, Denoised10 min

About this notebook

Brain Drip is written by one person, in long-form, with no schedule. The goal is to leave behind a small collection of explanations I would have wanted to read when I was learning these things.

The material assumes you can read code and follow a derivation, but it does not assume any specific background. Each Drip is self-contained.

How to read it

If you want a single concept explained well, open a Drip. If you want to learn an area properly, work through a Course. If you want to read what I’m thinking about, the Notes are the best place. If you want to build something this afternoon, start with a Blueprint.

Colophon

Set in Source Serif 4 and Inter Tight. Written and typeset in HTML. Source available on request. No ads, no tracking, no email gate.

© 2026 Brain Drip.