AI visionary Andrei Karpathyco-founder of OpenAI And Eureka Labs boss, just exceeded expectations with his latest open-source gem, nanochat. In a candid While developers hunt”Vibe coding“Dreams, Karpathy’s nanochat stands tall as a testament to human courage, with a whopping 8,000 lines of pure, handcrafted code. This release builds on his hit nanoGPT project, which moves from mere pre-training to a complete training and inference that anyone can adapt and perform. In an age where AI tools promise to code for you, Nanochat highlights the raw power of manual control and provokes cheers and debate in tech circles.
Nanochat Pure Code Roots: Specifications, Issues and AI Wake-Up Call
Karpathy didn’t mince his words when a curious follower probed his coding process. “It’s actually completely handwritten (with tab autofill),” he responded to X. He even tested AI agents like Claude And Codexbut quickly threw them away. “They just didn’t work well enough at all and weren’t helpful, possibly the repository is too far away from the data distribution,” he explained. This nanochat revelation caused a stir among AI skeptics Gary Marcus I’ll jump in to hail it as evidence of persistent AI failures. “Andrej Karpathy, inventor of the term vibration coding, hand coding. And he confirms – once again – that current AI has not solved the distribution shift,” Marcus tweeted, continuing his decades-long crusade against AI over-promise since 1998.
For tinkerers looking into nanochat, here’s what this powerhouse includes:
- Code volume: Approximately 8,000 lines spanning the entire pipeline from data preparation to deployment.
- Core focus: Trains and infers a simple ChatGPT-style model, going beyond the pretraining limits of nanoGPT.
- Tech stack: Relies on standard tools such as PyTorch; no fancy AI tools – just clean, efficient Python.
- Accessibility: Open-source on GitHub, inviting forks for custom bots or research tweaks.
Also read: SEO Explained: A Step-by-Step Guide to Search Engine Optimization
Karpathy’s move changes the script on AI dependency and urges programmers to hone their skills amid the temptations of tools. As nanochat gains traction — the forks are already peaking — it’s driving SEO searches for “handwritten AI code” and “nanochat tutorial.” Will this spark a backlash against lazy AI coding? Or prove that people still control the repository? Tech watchers are betting on the latter, with Karpathy taking the lead back to basics.
More news to read: AI layoffs on a large scale: Company faces setback despite funding
Walmart and OpenAI partner for ChatGPT Shopping Function
#Andrej #Karpathy #reveals #Nanochat #handwritten