# karpathy/autoresearch

AI agents running research on single-GPU nanochat training automatically

Repository: https://github.com/karpathy/autoresearch
Canonical: https://ross.abutalabs.com/products/autoresearch
Language: Python
License Family: other
Last push: 2026-03-26T00:07:37+00:00

## Health v2 (maintenance only)
Score: 48/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 74, release rhythm 35, longevity 12
- inputs: {"age_days": 180, "days_push": 161, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 94753, forks 13361 (observed 2026-08-28T04:12:23.530720+00:00)

## What it is
A framework by Andrej Karpathy where AI agents autonomously run LLM research experiments on a single-GPU nanochat training setup. The agent edits train.py, trains for a fixed 5-minute budget, evaluates val_bpb, and iterates overnight based on program.md instructions.

## Use cases
- run autonomous AI agent experiments on LLM training
- let an AI agent improve a GPT training loop overnight
- automate machine learning research experiments on a single GPU
- experiment with agent-driven code modification and evaluation loops
- set up an autonomous AI research organization with program.md
- benchmark architecture and optimizer changes fairly with val_bpb

## When to choose
- you have a single NVIDIA GPU and want to explore autonomous AI-driven ML research
- you want a minimal, hackable baseline for agent-run training experiments
- you want to learn how LLM training works and watch agents iterate on it

## When to avoid
- you need production LLM training at scale or multi-node clusters
- you have no NVIDIA GPU available
- you need a stable, licensed dependency for production systems (no license is provided)

## Facets
- artifact type: framework
- maturity: active
- function: agent-framework, llm-training, machine-learning, prompt-engineering
- domain: artificial-intelligence, large-language-models, machine-learning, deep-learning
- platform: python, cli
- tags: autonomous-research, nanochat, single-gpu, ai-agents, experiment-automation, self-improving-code, karpathy, gpu, linux

## Member repositories
- karpathy/autoresearch (main) score 48

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:12:23.530720+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T16:12:43.530910+00:00, confidence not recorded.
  - readme: https://github.com/karpathy/autoresearch (fetched 2026-08-28T04:12:23.530720+00:00, sha 304a04deaa68)
- Data as of 2026-08-30T08:39:29.467469+00:00.