Andrej Karpathy
Founder, Eureka Labs
'Software 2.0' — the claim that neural networks are replacing hand-written code — was Andrej Karpathy's phrase before it became industry consensus. An OpenAI co-founder and former Tesla AI director, he taught a generation of engineers through nanoGPT and nanochat, ran the education startup Eureka Labs for about 22 months, then joined Anthropic's pretraining team in May 2026.
Why they matter
No one else moves between building frontier models and explaining them at Karpathy's level: his from-scratch courses and nano repositories — over 120,000 GitHub stars combined — function as the field's de facto curriculum, and 'vibe coding,' a term he tossed off, became the name for an entire style of AI-assisted programming. His choice to fold Eureka Labs' independence into Anthropic's pretraining team is a signal about where frontier leverage now sits.
Leadership intelligence
TECHi profile brief
Andrej Karpathy is an OpenAI co-founder and former Tesla director of AI who joined Anthropic's pretraining team in May 2026, working under Nick Joseph on the large-scale training runs behind the Claude models. He previously ran Eureka Labs, an AI-native education startup, for about 22 months, and built widely used teaching resources including the nanoGPT and nanochat open-source repositories. Nanochat, released in October 2025, implements a complete ChatGPT-style training pipeline in a single hackable codebase.
TECHi thesis
Karpathy's career is a repeated trade between reach and access — and he keeps choosing access when the frontier heats up. He left Tesla's AI directorship, then OpenAI, then his own Eureka Labs, each time concluding the most formative work was happening somewhere else. Joining Anthropic's pretraining team as a hands-on researcher, rather than taking a chief-scientist title anywhere, says he believes the next few years of LLM progress will be decided in training runs, not in products or pedagogy. The bet cuts against his own Software 2.0 gospel of democratized capability: the man who taught everyone to build a ChatGPT clone on a shoestring now works where nine-figure training runs happen. If pretraining gains keep compounding, he picked right; if progress shifts decisively to post-training and agents, the educator rejoined the wrong layer.
Distribution leverage
84/100Karpathy's distribution is an audience, not an org chart: a nano-series with 120,000-plus GitHub stars and lectures that shape how new ML engineers learn mean his technical opinions propagate faster than most companies' launches.
Execution record
80/100Shipped at every layer — Tesla's production vision stack, OpenAI's founding research, teaching tools people actually finish. Eureka Labs ending without a flagship product is the one incomplete on the record.
Audience demand
76/100Interest spikes around every repo release and job change; the Anthropic move drew mainstream coverage usually reserved for CEO transitions — rare for what is, on paper, an individual-contributor role.
Career map
Inflection points
Origin
Software 2.0 framing
Articulated Software 2.0 while running Tesla's Autopilot AI, giving the industry its vocabulary for neural networks as a replacement for hand-written code.
Scale
AI education
Turned frontier-lab knowledge into open teaching infrastructure — nanoGPT and nanochat became the standard way engineers learn to build LLMs from scratch.
Watch
Eureka Labs
What his pretraining work at Anthropic yields — and whether he returns to the education mission he says he still plans to resume.
Search answers
People ask about Andrej Karpathy
Who is Andrej Karpathy?
Andrej Karpathy is an AI researcher on Anthropic's pretraining team, an OpenAI co-founder, and Tesla's former director of AI, where he led the Autopilot computer-vision effort. Between those roles he became the field's best-known teacher through open lectures and minimal from-scratch code repositories.
Why is Andrej Karpathy important in AI?
Because he compresses frontier practice into public knowledge. His nano-series repositories — nanoGPT, then nanochat's full ChatGPT-style pipeline in one codebase — have accumulated over 120,000 GitHub stars and serve as the field's de facto curriculum. His Software 2.0 essay predicted, years early, that neural networks would displace hand-written code.
What is Andrej Karpathy known for?
Software 2.0, the nanoGPT and nanochat repositories, coining 'vibe coding,' founding Eureka Labs, and his earlier runs as OpenAI co-founder and Tesla AI director. Since May 2026 he has worked on Claude pretraining at Anthropic.
Is Andrej Karpathy connected to a public company?
Not currently. Anthropic is private — its Series H valued it at $965 billion — and Eureka Labs never sold shares publicly. His clearest public-market association is historical: he previously ran AI at Tesla (TSLA).