Karpathy's "autoresearch" broke the internet

By Greg Isenberg

Categories: Startup, Product

Summary

Andrej Karpathy's AutoResearch is an AI agent that runs continuous ML experiments autonomously, testing hypotheses and keeping only winning configurations—enabling founders to optimize code, products, and business decisions 24/7 without manual intervention.

Key Takeaways

  1. AutoResearch operates as a closed feedback loop: set goal → AI plans experiment → edits code/settings → runs 5-minute GPU training → reads metrics → discards failed configs, saves winners → repeats until optimization complete.
  2. Requires NVIDIA GPU hardware or cloud access—cannot run on MacBook M1 or consumer chips. This is a critical infrastructure requirement before attempting to implement.
  3. Works beyond ML: optimize any software piece by creating a program.md markdown file and benchmark script, then let the agent iterate. Shopify CEO Toby validated this approach for production software optimization.
  4. Define success metrics explicitly (cheaper leads, more clicks, higher sales, better model accuracy) so the AI knows what 'better' means and can autonomously optimize toward that single objective.
  5. Mental model: treat AutoResearch as a delegated research boss with access to code, GPU, and internet. It runs planning-acting-reading loops continuously, logging all experiments, charts, metrics, and delivering summaries in natural language.

Topics

Transcript Excerpt

Andre Carpathy, I mean, one of the godfathers of AI, has just launched something called auto research. And auto research is a huge deal, and it's going viral on Twitter. And I just wanted to do an episode where I can explain to you in the clearest way possible what it is, what are the use cases, how to make money from it, how to be more productive with it, how to create impact with it. And by the end of this episode, I'm going to give you a bunch of different ideas, use cases for...