Reported by sourceNew
Let a coding agent run overnight ML training experiments
Point a coding agent at autoresearch so it edits a small GPT training script, runs short timed experiments and keeps what improves.
karpathy/
AI agents running research on single-GPU nanochat training automatically
What it does
AI agents running research on single-GPU nanochat training automatically
Summary based on the GitHub description (opens in a new tab)
What the README covers
- autoresearch
- How it works
- Quick start
- Running the agent
- Project structure
- Design choices
- Platform support
- Notable forks
- License
Getting started
Excerpt from the README · Checked 15 Oct 2026
Open the README and explore its “Quick start” section. Follow its linked resources for the project-specific workflow.
Shown exactly as published. Top 200 Repos has not run or tested it. View in README (opens in a new tab)
Ranking trend
Global rank in each recorded snapshot. A lower number is a higher rank.
From #137 to #143 across 2 snapshots since 15 Oct 2026.
What people build with it
Reported by sourceNew
Run unattended ML training experiments with autoresearch
Red Hat's Roy Belio containerized Karpathy's autoresearch on OpenShift AI with H100 GPUs and let Claude Code run 198 training experiments over 24 hours.