Karpathy Project Taxonomy

This taxonomy groups Karpathy public corpus items by the role they play in AI learning, implementation, and research tooling.

Snapshot

  • Crawl date: 2026-05-16
  • Total items: 172
  • GitHub repositories: 63
  • New entries this run: 0
  • Changed entries this run: 1

Browser / JavaScript ML experiments

  • Items: 2

  • GitHub repositories: 2

  • calorie - github-repository, summary-only

  • tsnejs - github-repository, summary-only

LLM education / from-scratch pedagogy

  • Items: 1

  • GitHub repositories: 1

  • cryptos - github-repository, summary-only

LLM training and inference systems

  • Items: 17

  • GitHub repositories: 17

  • build-nanogpt - github-repository, summary-only

  • cpython - github-repository, summary-only

  • examples - github-repository, partial excerpt

  • hn-time-capsule - github-repository, summary-only

  • karpathy.github.io - github-repository, summary-only

  • llama2.c - github-repository, partial excerpt

  • llm.c - github-repository, partial excerpt

  • LLM101n - github-repository, summary-only

  • minbpe - github-repository, partial excerpt

  • minGPT - github-repository, partial excerpt

  • nanochat - github-repository, partial excerpt

  • nanoGPT - github-repository, partial excerpt

  • ng-video-lecture - github-repository, summary-only

  • randomfun - github-repository, summary-only

  • reader3 - github-repository, summary-only

  • rendergit - github-repository, summary-only

  • transformers - github-repository, partial excerpt

Minimal implementations

Neural network fundamentals

Personal heuristics / AI philosophy / learning advice

Research automation / agentic science

  • Items: 6

  • GitHub repositories: 6

  • autoresearch - github-repository, summary-only

  • jobs - github-repository, summary-only

  • llm-council - github-repository, summary-only

  • reinforcejs - github-repository, summary-only

  • scriptsbots - github-repository, summary-only

  • tf-agent - github-repository, summary-only

Research tooling and paper workflows

Talks, courses, and videos

Tokenization and language modeling

  • Items: 1

  • GitHub repositories: 1

  • rustbpe - github-repository, partial excerpt

Vision / multimodal / captioning

  • Items: 5

  • GitHub repositories: 5

  • forestjs - github-repository, partial excerpt

  • nn - github-repository, summary-only

  • simple-amt - github-repository, partial excerpt

  • svmjs - github-repository, partial excerpt

  • ulogme - github-repository, summary-only

Counterpoints And Gaps

  • Generated categories are heuristic and should be refined when a repo’s README or file tree shows a stronger interpretation.
  • Some repositories are forks, experiments, or old teaching artifacts; popularity should not be mistaken for current endorsement.
  • Source code is summarized through file trees and metadata unless public redistribution is clearly license-safe.