Andrej Karpathy
Andrej Karpathy (born 23 October 1986) is a Slovak-Canadian computer scientist and artificial intelligence researcher. He is a founding member of OpenAI, served as Director of AI at Tesla leading the Autopilot computer vision team, and founded the AI education company Eureka Labs.[^c1][^c3][^c4] He is known for his contributions to deep learning and computer vision, his educational work through Stanford's CS231n course and the "Zero to Hero" YouTube series, and for coining the terms "Software 2.0", "vibe coding", "Software 3.0", "agentic engineering", and "AI psychosis."[^c2][^c7][^c8]
Karpathy earned a BSc in computer science and physics from the University of Toronto in 2009, an MSc from the University of British Columbia in 2011, and a PhD from Stanford University in 2015 under the supervision of Fei-Fei Li. His research focused on the intersection of natural language processing and computer vision. During his undergraduate years, he gained recognition in the speedcubing community for his CFOP tutorial videos on YouTube under the alias badmephisto.[^c9] At Stanford, he created and taught CS231n, the university's first deep learning course, which grew from 150 to 750 students and became one of the largest classes on campus.[^c6]
In 2024, Karpathy founded Eureka Labs, an AI education platform built on a philosophy of "Teacher + AI symbiosis," where human educators design curriculum and AI teaching assistants deliver personalized instruction at scale.[^c5] He continues to produce educational content through his YouTube channel and maintains numerous open-source projects including micrograd, makemore, nanoGPT, arxiv-sanity-lite, and autoresearch.
In 2025 and 2026 he delivered major keynote addresses at Y Combinator's AI Startup School and Sequoia Capital's AI Ascent conference, where he introduced the Software 3.0 paradigm and related concepts including agentic engineering, jagged intelligence, and verifiability.[^c7] In February 2026, he formally declared that vibe coding had been superseded by agentic engineering as the discipline for professional software development, describing the new paradigm as one where developers no longer write code directly but instead command AI agents with precision and architectural intent.[^c10]
In April 2026, Karpathy introduced the [[concepts/llm-wiki|LLM Wiki]] paradigm for AI-maintained personal knowledge bases, arguing that LLMs could eliminate the maintenance burden that causes knowledge bases to collapse.[^c11] The following month he outlined a roadmap for AI output formats from plain text and Markdown through HTML to interactive neural video — video generated directly by diffusion models without intermediate code.[^c12] His open-source autoresearch project was also expanded by the community into a decentralized swarm of 81 AI agents that spontaneously self-organized into specialized research roles and developed an emergent peer review system.
Note: course creator attribution
index.md describes Karpathy as the one who "created and taught CS231n," while [[projects/cs231n]] and [[people/andrej-karpathy]] both state the course was designed "together with" his advisor Fei-Fei Li. The index page foregrounds Karpathy's role as primary instructor and originator — consistent with it being his biographical page — while the dedicated course page gives fuller co-creator credit.