microgpt
Source
- Source kind:
blog-rss-item - URL: http://karpathy.github.io/2026/02/12/microgpt/
- Discovery source: https://karpathy.github.io/feed.xml
- License:
NOASSERTION - Distribution policy:
public-summary-local-archive-only - Public mirror status:
summary-only - Content hash:
ca1aaa9f818b3e97e45c8b18c557a04a2714d496e77b68ca9f0e74cb08f12020 - First seen: 2026-05-15
- Last changed: 2026-05-15
Classification
- Primary category: Personal heuristics / AI philosophy / learning advice
- Corpus source note: 2026-05-15-karpathy-public-corpus
- Project taxonomy: karpathy-project-taxonomy
- Idea map: karpathy-idea-map
- Topic hub: karpathy-public-work
Summary
This is a brief guide to my new art project microgpt , a single file of 200 lines of pure Python with no dependencies that trains and inferences a GPT. This file contains the full algorithmic content of what is needed: dataset of documents, tokenizer, autograd engine, a GPT-2-like neural network architecture, the Adam optimizer, training loop, and inference loop. Everything else is just efficiency. I cannot simplify this any further. This script is the culmination of multiple projects (micrograd, makemore, nanogpt, etc.) and a decade-long obsession to simplify LLMs to their bare essentials, and I think it is beautiful ð¥¹. It even breaks perfectly across 3 columns: Where to find it: This GitHub gist has the fu…
What This Teaches
- How Karpathy frames technical judgment, learning, research, or AI systems in long-form prose.
- Useful as a high-signal idea source for research taste, project framing, and agent workflow design.
Why It Matters
Karpathy’s posts often crystallize reusable heuristics; this wiki should preserve the ideas without relying on chat memory.
Public Handling Notes
- EXTRACTED: This page records public metadata and a source-grounded summary.
- INFERRED: Full local preservation, when available, is for private/local use unless a license or explicit source policy makes public redistribution safe.
- Do not treat this page as permission to republish unlicensed source text or code wholesale.