Who is Andrej Karpathy?
Andrej Karpathy is a researcher, engineer, and educator widely regarded as one of the clearest explainers of modern AI. He earned his PhD at Stanford under Fei-Fei Li, working on connecting images and language, then became a founding member of OpenAI. He later led the Autopilot vision team at Tesla as director of AI, returned briefly to OpenAI, and in 2024 launched Eureka Labs, a startup focused on AI-native education. Beyond his industry roles, Karpathy is known for a body of free teaching material that has trained a generation of practitioners, including his “Neural Networks: Zero to Hero” video course and small, readable codebases such as micrograd and nanoGPT. He coined the influential phrases “Software 2.0” and, more recently, “Software 3.0” to describe how machine learning is reshaping the practice of programming. His blog and talks are treated as reference points across the field.
What does Andrej Karpathy think about AI?
Karpathy is broadly optimistic but pointedly anti-hype. He frames large language models less as looming agents and more as a new kind of computer and a powerful tool that amplifies ordinary people. In his essay “Power to the people” he argues that LLMs reverse the usual top-down path of transformative technology, reaching individuals first and corporations and governments only later, which he sees as unusually empowering.
His “Software 3.0” framing holds that natural language prompts are becoming a form of source code, with neural networks as the runtime, building on his earlier “Software 2.0” idea that models are programs compiled from data rather than written by hand. At the same time he is a noted realist about timelines. He has pushed back on breathless claims, describing the coming period as the “decade of agents” rather than the “year of agents,” and noting that today’s systems lack the reliability, multimodality, and continual learning needed to act autonomously. He emphasises what humans still own: judgment, taste, verification, and the work of turning flawed model output into something trustworthy.
What is Andrej Karpathy’s role in the AI race?
Karpathy’s influence comes less from running a frontier lab today and more from shaping how the field thinks and learns. As a founding member of OpenAI and the leader who built much of Tesla’s Autopilot vision stack, he sat at the centre of two of the most consequential AI efforts of the past decade. His decision to leave large companies and start Eureka Labs signals a bet that the bottleneck in the AI race is increasingly people and understanding, not just compute. His talks, including the widely viewed Y Combinator AI Startup School keynote on Software 3.0, set vocabulary that practitioners and executives then adopt. His tutorials lower the barrier to entry, expanding the pool of people who can build with these systems. He also functions as a respected reality check: when he says agents need a decade to mature, it tempers expectations across the industry. In a race often driven by hype cycles, Karpathy’s role is to clarify what is real, what is hard, and what is still science fiction.
Where does Andrej Karpathy work?
Eureka Labs, which Karpathy announced in 2024, is an AI-native education company. Its premise is that great teaching can be paired with AI to make high-quality learning radically more accessible, with human-designed course material supported by AI teaching assistants that help students work through it. The first planned offering is an undergraduate-level course on building AI, reflecting Karpathy’s long-running focus on demystifying neural networks. The company is small and early-stage compared with the frontier labs, and Karpathy has framed it as a long-term project rather than a sprint. It extends the work he had already been doing for free through his Zero to Hero course and open codebases, now organised into a venture aimed at scaling that approach to far more learners.
What are Andrej Karpathy’s key projects?
Karpathy’s projects are unusually influential for how compact they are. micrograd is a tiny automatic differentiation engine that teaches backpropagation in a few hundred lines. nanoGPT is a minimal, hackable implementation of a GPT-style model that many people use to learn how transformers work and to run their own experiments. makemore is a character-level language model used for teaching. His “Neural Networks: Zero to Hero” YouTube course builds these ideas up from first principles, and remains one of the most recommended free routes into deep learning. Earlier, his Stanford course CS231n on convolutional networks shaped how computer vision was taught. Through Eureka Labs he is now packaging this teaching philosophy into structured courses. Alongside the code, his essays and talks, Software 2.0, Software 3.0, and Power to the people, function as conceptual projects that frame how the field describes itself.
What has Andrej Karpathy written about AI?
A selection of Karpathy’s most influential essays, courses, and open-source projects:
- Software 2.0. The essay arguing neural networks are programs compiled from data, 2017.
- Power to the people: How LLMs flip the script on technology diffusion. On AI reaching individuals first, 2025.
- Neural Networks: Zero to Hero. His free, build-from-scratch deep learning video course.
- nanoGPT. A minimal, hackable GPT implementation for learning and research.
- micrograd. A tiny autograd engine that teaches backpropagation.
- nn-zero-to-hero. The companion code repository for the Zero to Hero course.
- Eureka Labs. His AI-native education company and its launch description, 2024.
Does Andrej Karpathy think humanity will survive AI?
StrideNote’s reading of how strongly their public work backs humanity coming through the AI transition, scored out of 10.
7 / 10. Karpathy scores well because his public work pulls steadily toward human understanding and agency. He treats AI as a tool that empowers individuals, insists humans retain judgment and verification, and actively widens access through free education. His “decade of agents” realism cools dangerous overconfidence about autonomous systems. He does not, however, center catastrophic or existential risk the way some peers do, and his framing is more about productivity and capability than about formal safety guarantees. That keeps him a notch below the very top. Still, by making the technology legible to millions and resisting hype, he strengthens the odds that society adapts deliberately rather than blindly, which is why his body of work reads as a clear positive for navigating the transition.
Is Andrej Karpathy a transhumanist?
StrideNote’s reading of how far they embrace transhumanism, the use of technology to transcend human biological limits through enhancement, longevity, or merging with machines, scored out of 10.
3 / 10. Karpathy shows little public interest in transhumanism. He treats AI as a tool that empowers people rather than a path to transcending biology, and rarely discusses enhancement, longevity, or merging with machines. This low ranking reflects an absence of stated transhumanist ambition rather than opposition to it.