Andrej Karpathy, co-founder of OpenAI and founder of Eureka Labs, revealed his latest open-source project, nanochat, was written entirely by hand, confirming that he did not rely on AI tools for the code.
In a post on X on October 13, Karpathy announced the release of nanochat, an open-source project that provides a full-stack training and inference pipeline for a simple ChatGPT-style model. The repository follows his earlier project, nanoGPT, which focused only on pretraining.
The repo consists of about 8,000 lines of code and covers the entire pipeline.
When asked by a user how much of the code was written manually, Karpathy replied, “Good question, it’s basically entirely hand-written (with tab autocomplete). I tried to use Claude/Codex agents a few times, but they just didn't work well enough at all and net unhelpful, possibly the repo is too far off the data distribution.”
The statement drew attention from AI researcher Gary Marcus, who noted that Karpathy’s comment reinforces ongoing challenges in AI generalisation. “Andrej Karpathy, inventor of the term vibe coding, hand-coding. And confirming — yet again — that current AI has not solved distribution shift (the core problem that I have been harping on since 1998),” Marcus wrote on X.
- Dispatch Network Raises ₹5 Cr Pre-Seed Led by Antler India to Build AI-Powered Delivery Network
- Coforge Launches BridgeCX With Pega, Mastercard for Customer Personalisation
- HCLTech Launches AI Agents to Support Mainframe Operations
- MIT Researchers Show LLMs Can Read and Edit Neural Network Weights Without Retraining
- Anthropic Posts $17 Bn in Q3 Revenue, Swings to Operating Loss
- OpenAI Signs Four Agreements With Abu Dhabi Across Government, Energy And Investment