Karpathy's 630-Line Agent Self-Loop Experiment
FinanceYF5 · x · 2026-07-07
In March 2026, Karpathy released a 630-line project: the Agent modifies code itself, trains for 5 minutes, and keeps the changes if metrics improve, or rolls back otherwise. Over two days, it ran 700 experiments and found 20 optimizations that he hadn't manually discovered in 20 years. Shopify's CEO ran it overnight, achieving a 19% quality improvement while halving the model size.
Related event: The Karpathy Loop: Building Autonomous AI Agent Workflows(9 posts)→
More from coding & agent
- OpenAI ships GPT-6.1 Terra; Brockman says its computer use now beats his own speed — gdb · 2026-10-27
- Kolab Benchmark Uses Pro Screen Recordings to Show AI Agents Overestimate Collaboration Quality — minnesotanlp · 2026-10-27
- Dev argues Claude burns far more tokens than GPT models for the same coding work — haider1 · 2026-10-27
- Stop building agents that can't change models: stable contracts, external state, real evals — rseroter · 2026-10-27
- Cheap models can cost more in agent harnesses than frontier ones, researcher warns — vasuman · 2026-10-27
- Kanzo: A 5k-Line Rust Kanban Binary With MCP for Claude Code and Codex — allmudi · 2026-10-27