Karpathy's "autoresearch" broke the internet
Greg Isenberg explores Auto Research, Andrej Karpathy's viral AI tool that automates experimental loops to optimize code, products, and business processes. The episode breaks down what Auto Research does, presents ten concrete business ideas for monetizing it, and provides a practical guide to getting started with the platform. Isenberg argues this represents a significant shift in how businesses can iterate and optimize at scale, comparing it to having a "super nerd robot intern" running experiments 24/7.
Key takeaways
- • Auto Research is an AI agent that autonomously plans experiments, edits code, runs training iterations, evaluates results, and keeps only winning configurations without human intervention.
- • Running Auto Research requires an Nvidia GPU (not compatible with MacBook M1), but users can rent cloud GPUs affordably from services like Lambda Labs, Vast AI, RunPod, or Google Colab.
- • Ten monetizable use cases exist: niche agents-in-a-box, A/B testing for landing pages/ads, research-as-a-service, embedded optimization features in existing SaaS products, optimization agencies, trading signal generation, lead qualification automation, finance operations automation, internal productivity labs, and done-for-you due diligence.
- • The simplest way to get started is using Claude Code to help install Auto Research and then running it on Google Colab with a T4 GPU, which requires minimal technical setup beyond pasting commands.
- • AgentHub, Karpathy's companion project, is a GitHub-like collaboration platform designed for agent swarms to work together on shared codebases, suggesting a future of multi-agent coordination.
- • The key to success is defining clear business goals (e.g., "more sales," "lower CAC," "higher test scores") and letting the agent iterate on solutions while maintaining human oversight, especially in high-stakes domains like trading or healthcare.
Recommendations (4)
"In my case, I don't have an Nvidia GPU. I'm actually using a MacBook and an M1 Pro. I know I need an upgrade to a new Mac."
Greg Isenberg · ▶ 20:59
"Auto Research is a huge deal, and it's going viral on Twitter. 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..."
Greg Isenberg · ▶ 0:06
"To get started, I'd recommend just tell Claude code to get you started. I went ahead and I basically gave Cloud Code the GitHub repo and was like I need help installing auto research by Karpathy."
Greg Isenberg · ▶ 19:53
"I personally use Google Collab. Why? I just know Google the best and trust Google the best. The easiest way to get started, you go to collab.google.com."
Greg Isenberg · ▶ 21:37
Mentioned (7)
More from these creators
23 AI Trends keeping me up at night
Stop Vibe Coding. Start Getting Customers.
Paperclip: Hire AI Agents Like Employees (Live Demo)
Firecrawl AI clearly explained (and how to make $$)
I fixed OpenClaw so it actually works (full setup)
Building AI Agents that actually work (Full Course)