Doug Does

AI systems · research · teaching · experiments

An internet workshop by Doug Schonholtz

Things I’m trying to understand by building them.

Practical research, courses, and small experiments about software, data, and artificial intelligence.

Current work

Everything has a home. Nothing has to be finished.

  1. 01 Course · in development LLMs from First Principles Interactive courses for software engineers: build a small language model, fine-tune a model for SQL, and explore the mathematics behind training. Enter course Supplementary Class 1 presentation
  2. 02 Research · interactive report The Entry-Level CS Job Market An explorable model of degree supply, job openings, and the assumptions hiding inside “entry level.” Explore report
  3. 03 Experiment · game design Glimmerroot, the Waywarden A playable character pitch with a browser prototype, original tile art, and a downloadable 3D handoff. Play experiment
  4. 04 Experiment · Fat Bear Week Fat Bear Fan Club Choose your top three, give a bear a nickname, and explore the stories of river legends and very good cubs. Meet the bears
  5. 05 Field notes · local AI Jev vs. Laya vs. Kev: same questions, different answers. Laya, Kev, and now Open-Jev take on the same four scenarios. Read the answers, probabilities, and timing methods. Read the post
  6. 06Research · reproducible benchmarkJev decision-model benchmarksFive checkpoints, 231 public decisions, and 100 private creative judge questions. Compare results and inspect every public answer.Explore results
  7. 07 Writing · Substack Notes from the workbench Field reports, longer arguments, and useful scraps from the projects above. Read the blog

Learning guides

Choose a resource. Plan the practice.

  1. 01Guide · AI learningSix ways to learn AICompare prerequisites, practice and depth across Karpathy, 3Blue1Brown, fast.ai, Hugging Face, Andrew Ng and Doug Does AI.Read the guide
  2. 02Study plan · Zero to HeroA study companion to Karpathy’s Zero to HeroPlan the eight-video sequence and use four checkpoints to test your understanding between coding sessions.Read the study plan
What connects it all

Build the thing. Measure the thing. Explain what changed.

I’m an AI engineer interested in how probabilistic systems become dependable software. This site is the public notebook: serious research next to teaching materials and whatever odd prototype seemed worth making that week.