Fetching from the wire…
Public story · 2026-07-21 · high
Karpathy, who joined Anthropic's pretraining team in May, says a personal knowledge base beats using AI to write code.
Why now: The post and the nano-series star count are being read together in the July 21 roundup, pairing Karpathy's advice with the clearest evidence for it.
Karpathy told developers to stop using AI to write code and start building a personal knowledge base with it. The post reached 21 million views on X.
Separately, Karpathy's nano-series, nanoGPT, nanochat, and micrograd, crossed 120,000 combined stars, per coverage of the series. That's ahead of most production ML frameworks in developer attention, for repos built to be read in an afternoon, not deployed at scale. For anyone shipping developer tools, that's the more durable signal than the view count.
Karpathy's Software 2.0 thesis shaped how a whole generation of engineers think about machine learning systems as learned code. He joined Anthropic's pretraining team in May. Coming from him, the post reads like a deliberate correction to the codegen-maximalist culture he helped build, not an outsider's complaint.
The 21 million figure comes from tracking a single post. Treat it as an approximate amplification number, not a verified reach metric.
Each link below shares sources, entities, or timing with this story.
Andrej Karpathy reacting to a product launch is usually worth more than the launch's own marketing, and his take on Anthropic's June 23 Claude Tag release is a genuinely useful frame for anyone building agents. His argument: LLM interaction has entered a third paradigm (Karpat...
Andrej Karpathy stood up at Sequoia Capital's AI Ascent 2026 and said what a lot of us have been thinking but hadn't articulated this cleanly. He called the current era "Software 3.0," defined as prompting an LLM interpreter, and declared December 2025 the tipping point when a...
1. Flip your multi-model pipeline to review-then-generate. Instead of using a reasoning model to plan before code generation, let the specialist generate freely and use reasoning tokens for review. Paper shows 90.2% pass@1 vs 87.2% for the planning pattern. Source 2. Audit you...
poetiq.ai — Detailed technical breakdown of how a $40K-hardware startup achieved 54% on ARC-AGI-2 (vs. Google's 45% at nearly 3x the cost). The key innovation: "learned test-time reasoning" — an iterative refinement meta-system where solutions are generated, receive structured...
19. Karpathy — microGPT 20. TechCrunch — Altman vs Anthropic 21. MIT Tech Review — LeCun AMI Labs 22. Dario Amodei — Adolescence essay 23. Simon Willison — Showboat and Rodney 24. Lenny's Newsletter — v0 25. ARC Prize — ARC-AGI-3
The decision, covered by Simon Willison, frames the ban as strategic: maintainers invest time reviewing contributions to mentor developers into trusted long-term contributors. If an LLM wrote the code, that mentorship is wasted. The wrinkle: Bun (acquired by Anthropic) runs a...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.