Skip to main content
← Back to sources

Andrej Karpathy Describes Autonomous AI Agent Loop Running 700 Experiments in Two Days

Published 2026-03-17Ingested 2026-03-25Agentic AIHigh

Summary

Fortune reports on a concept being called 'The Karpathy Loop,' referencing former Tesla and OpenAI researcher Andrej Karpathy's description of autonomous AI agents capable of running approximately 700 experiments over just two days. The piece examines how this rapid, agent-driven iteration cycle represents a potential inflection point in how AI systems conduct research and engineering tasks, with agents autonomously designing, executing, and evaluating experiments at a pace far beyond human capa

Alignment: Reinforces current position
Related Positions: agentic-workflows.md, ai-assisted-development-tooling.md, enterprise-ai-delivery.md
Related Partnerships: cognition-windsurf-devin.md, anthropic-claude.md
karpathy-loopagentic-aiautonomous-agentsai-experimentationagentic-workflowsai-engineeringautonomous-codingrapid-iterationai-native-developmentagent-orchestration
Andrej Karpathy Describes Autonomous AI Agent Loop Running 700 Experiments in Two Days — Intelligence — Agentic Developer Tools Radar · Signal