Skip to Content
About This Adaptation

About This Adaptation

🌐 English · Русский · Eesti

This site is an adapted edition of microgpt-3d-tutorial  by lxb12123 , an interactive 3D tutorial for Andrej Karpathy’s microGPT . The original is published under the MIT license; this adaptation keeps that license and is maintained by Tyche Institute  (Estonian non-profit, MTÜ) in its fork .

What is original work

Everything that makes this tutorial run is the upstream authors’ craft:

  • the five lessons’ core text, structure, and annotated code walkthroughs;
  • every 3D sandbox — the React Three Fiber scenes, the Blender-authored assets, the timeline and HUD machinery;
  • the TypeScript port of microGPT that computes every logit, gradient, and attention weight live in your browser;
  • and, underneath it all, Andrej Karpathy’s ~150-line pure-Python microGPT itself.

What Tyche Institute added

The adaptation layer is aimed at junior AI researchers — anyone taking their first serious steps into how language models actually work:

  • 00 · Foundations — a prerequisites lesson covering the five ideas the tutorial builds on, assuming school math only;
  • Glossary — every term in one sentence;
  • “Before you dive in” primers at the top of each lesson, translating jargon into lesson-00 vocabulary;
  • “Try this” experiment lists for each sandbox;
  • narrative introductions to each lesson and the homepage;
  • complete Russian and Estonian translations of all of the above and of the original lesson texts.

Translations were drafted with AI assistance (Claude) under human editorial control, and are reviewed by Tyche Institute staff; the Estonian text gets a native-speaker pass. If you spot an error — in any language — please open an issue .

Why a research institute hosts a school tutorial

Tyche Institute works on verifiable evidence for AI systems. A tutorial where every claim about “how GPT works” can be checked against live computation in your own browser — no backend, no black box, weights and code in the open — is what that principle looks like in education. This adaptation is part of the Institute’s Lab  line and its research interest in attestable educational AI.

License

MIT — both the upstream project and this adaptation. See the repository  for details.

Last updated on