# Andrej Karpathy

> Andrej Karpathy is a Slovak-Canadian-American AI researcher and educator who co-founded OpenAI, led Tesla's Autopilot AI as Director of AI from 2017-2022, and now runs Eureka Labs, an AI-native education startup. He coined 'vibe coding' (Collins Dictionary 2025 Word of the Year), created the legendary Stanford CS 231n course, and builds open-source tools like nanoGPT that have become essential references for the global ML community. His YouTube 'Zero to Hero' series and multi-hour deep dives have demystified deep learning for millions.

- **Role:** CEO & Founder at Eureka Labs
- **Organizations:** Eureka Labs, OpenAI (co-founder), Tesla (former), Stanford University (former)
- **From:** Bratislava, Czechoslovakia (now Slovakia)
- **Nationality:** Slovak-Canadian-American
- **Education:** BSc in Computer Science & Physics (double major, minor in Mathematics), University of Toronto, MSc in Computer Science, University of British Columbia, PhD in Computer Science (advisor: Fei-Fei Li), Stanford University
- **Known for:** Co-founded OpenAI (2015), one of the most influential AI labs in history, Led Tesla Autopilot AI as Director of AI (2017-2022), deploying vision-only neural nets to 3+ million vehicles, Created CS 231n at Stanford - grew from 150 to 750+ students, now one of the most-watched AI courses globally

## Career timeline

- **1986** — Born in Bratislava, Czechoslovakia (now Slovakia) on October 23
- **2001** — Family moves to Toronto, Canada when he is 15
- **2005-2009** — Undergraduate at University of Toronto (CS & Physics double major)
- **2009-2011** — MSc at University of British Columbia
- **2011-2016** — PhD at Stanford under Fei-Fei Li; internships at Google Brain, Google Research, DeepMind
- **2015** — Created and taught CS 231n: Convolutional Neural Networks for Visual Recognition at Stanford
- **2015** — Co-founded OpenAI as founding research scientist
- **2017** — Joined Tesla as Director of AI and Autopilot Vision
- **2017** — Published influential 'Software 2.0' essay on Medium
- **2022** — Left Tesla in July after 5 years leading Autopilot AI
- **2023** — Rejoined OpenAI in February, focused on midtraining and synthetic data
- **2024** — Left OpenAI in February after exactly one year
- **2024** — Founded Eureka Labs in July - AI-native education platform
- **2025** — Coined 'vibe coding' in a viral February tweet; Collins Dictionary names it 2025 Word of the Year
- **2025** — Released 3h31m 'Deep Dive into LLMs like ChatGPT' on YouTube
- **2026** — Stopped writing code manually in December 2025; now fully orchestrates AI agents
- **2026** — Released microgpt - GPT training in 243 lines of pure Python, no PyTorch
- **2026** — AutoResearch runs 700 autonomous experiments, achieves 11% training speed improvement

## Achievements

- Co-founded OpenAI (2015), one of the most influential AI labs in history
- Led Tesla Autopilot AI as Director of AI (2017-2022), deploying vision-only neural nets to 3+ million vehicles
- Created CS 231n at Stanford - grew from 150 to 750+ students, now one of the most-watched AI courses globally
- nanoGPT: 57,500+ GitHub stars, the go-to reference implementation for GPT training
- Coined 'vibe coding' - Collins Dictionary 2025 Word of the Year
- MIT Technology Review 'Innovators Under 35' (2020)
- TIME Magazine '100 Most Influential People in AI' (2024)
- 30+ peer-reviewed papers with 50,000+ academic citations
- 180,000+ GitHub followers; combined 40,000+ stars across repositories
- 2.3 million Twitter/X followers
- Founded Eureka Labs (2024) to democratize world-class AI education
- Created micrograd - tiny autograd engine used by thousands to learn backpropagation from scratch
- llm.c: LLM training in pure C/CUDA - 29,800+ stars
- AutoResearch: AI agent ran 700 experiments in 2 days, discovering 11% training speedup
- 'badmephisto' Rubik's cube speedsolving channel - 9M+ views; content used by world record holder Feliks Zemdegs

## Latest updates

- **2026-04** — Presented at YC/Sequoia events on Software 3.0 and agentic engineering philosophy
- **2026-03** — AutoResearch AI agent ran 700 experiments autonomously, achieving 11% training speed improvement; announced he hasn't written code since December 2025
- **2026-02** — Released microgpt - GPT training in 243 lines of pure Python with no external dependencies
- **2025-12** — Released '2025 LLM Year in Review' video on YouTube
- **2025-12** — Crossed the 'Coherence Threshold' - stopped writing code manually, shifted to full AI agent orchestration
- **2025-02** — Coined 'vibe coding' in a viral tweet; Collins Dictionary later named it 2025 Word of the Year
- **2025-02** — Released 3h31m 'Deep Dive into LLMs like ChatGPT' YouTube video
- **2024-07** — Founded Eureka Labs, an AI-native education platform

## Links

- Website: https://karpathy.ai
- LinkedIn: https://www.linkedin.com/in/andrej-karpathy-9a650716/
- Twitter/X: https://x.com/karpathy
- GitHub: https://github.com/karpathy
- YouTube: https://www.youtube.com/@AndrejKarpathy
- Instagram: https://www.instagram.com/karpathy

---

Profile page: https://yespress.io/andrej-karpathy
Published by YesPress — https://yespress.io
Last updated: 2026-05-04