learning-with-llms
Interactive explainers for learning technical topics deeply — intuition, full mathematics, live simulations, hard quizzes, spaced repetition, and build-it-yourself labs. Built with Claude Code.
I use LLMs to teach myself new technical topics, and this site is where the results live: each topic ships as a complete learning system rather than a blog post.
Every topic includes:
- An interactive explainer — background → intuition → the real mathematics → the key papers, with live simulations you can poke, interactive figures, faded derivations, and a concept map.
- Hard quizzes — five per section, every option explained whether you get it right or wrong.
- Spaced repetition — a Leitner-box review deck in the browser, plus an Anki export.
- A one-page cheat sheet — every formula with its punchline.
- A lab — implement the core algorithms in numpy against a test suite that knows the right answers.
The first topic is diffusion language models — why frontier LLMs generate one token at a time, and how diffusion lets them paint the whole answer instead.
The whole pipeline — a spec-to-HTML renderer, numpy-computed Plotly figures, and the Claude Code skill that drives it — is open source at github.com/raghuramshankar/learning-with-llms.