Andrej Karpathy

Early Life and Education in Canada

Andrej Karpathy immigrated to Canada with his family at the age of 15, settling in Toronto. He pursued his undergraduate studies at the University of Toronto, where he earned dual bachelor's degrees in Computer Science and Physics in 2009. For his graduate work, Karpathy moved west to attend the University of British Columbia, completing a master's degree in 2011 with a focus on physically simulated figures. During his university years, he began creating educational content online, launching a YouTube channel under the name badmephisto in 2006. This channel initially gained popularity for its detailed Rubik's Cube solving tutorials, which were used by champion speedcubers like Feliks Zemdegs and have amassed millions of views.

Karpathy's academic journey culminated at Stanford University, where he earned a PhD in 2015 under the supervision of renowned AI researcher Fei-Fei Li. His doctoral research focused on the intersection of natural language processing and computer vision, developing deep learning models to bridge these two fields. This foundational work positioned him at the forefront of multimodal AI research. His time at Stanford also included a significant teaching contribution, as he authored and became the primary instructor for the university's first deep learning course, CS 231n: Convolutional Neural Networks for Visual Recognition, which grew from 150 students in 2015 to 750 by 2017.

Foundational Work at OpenAI and Tesla

In 2015, Andrej Karpathy became a founding member and research scientist at OpenAI, the artificial intelligence research laboratory. He contributed to the organization's early research in deep learning during a formative period for the field. His career took a major turn in June 2017 when he joined Tesla as its director of artificial intelligence, reporting directly to CEO Elon Musk. In this role, Karpathy led the Autopilot Vision team, which was responsible for developing the computer vision systems crucial for Tesla's advanced driver-assistance technology. His work involved applying deep neural networks to interpret visual data from vehicle cameras for real-time navigation and object recognition.

Karpathy's leadership at Tesla lasted nearly five years, during which he was recognized as one of MIT Technology Review's Innovators Under 35 in 2020. After a several-month sabbatical, he announced his departure from Tesla in July 2022. Following his exit, he dedicated time to creating detailed technical YouTube videos and blog posts that demystified the construction of artificial neural networks, further cementing his role as a public educator. In February 2023, he announced a return to OpenAI, but his second tenure lasted only a year, with an OpenAI spokesperson confirming his departure in February 2024.

Founding Eureka Labs and AI Education

In July 2024, Andrej Karpathy announced the founding of his own company, Eureka Labs, an AI education platform. The company's stated mission is to provide high-quality, accessible learning resources for understanding and working with large language models. Its first major product was the AI course titled LLM101n. Karpathy also spearheaded a broader educational series called "Zero to Hero," which offers foundational lessons on LLM concepts aimed at taking learners from beginner to advanced understanding. Eureka Labs has advocated for the use of AI teaching assistants to scale education, a concept that has attracted criticism over data privacy concerns and the potential loss of personal connection in learning.

The same year, Karpathy's influence in the AI field was formally recognized when he was named one of Time Magazine's 100 Most Influential People in AI. His educational work through Eureka Labs represents a strategic pivot from pure research and industry application to shaping how the next generation of developers and researchers learn core AI principles. The platform's materials, often shared via its official GitHub account, continue his long-standing practice of open and clear technical communication that began with his Stanford course and YouTube channel. This focus on education builds directly on his early success in explaining complex topics, first with Rubik's Cube tutorials and later with deep learning.

Thought Leadership and Coined Concepts

Andrej Karpathy has significantly influenced AI discourse through his public talks and writing. In February 2025, he coined the term "vibe coding" to describe a phenomenon where hobbyists use AI tools to build applications primarily by typing natural language prompts, often without deep programming knowledge. By 2026, his commentary evolved to describe a major shift in developer workflows. He publicly noted that his own coding process had rapidly changed from being 80 percent manual work with autocomplete to 80 percent agent-driven coding with only 20 percent manual edits, a transition he said happened around December 2025. He also humorously predicted that 2026 would be the year of the "Slopacolypse," referring to a potential influx of low-quality, AI-generated content and code.

In talks such as one at the Sequoia Ascent conference in 2026, Karpathy elaborated on frameworks for understanding AI's future, including Software 3.0, agentic engineering, verifiability, and jagged intelligence. He drew a clear distinction between simple "vibe coding" and more structured "agentic engineering," which involves careful specification design, difference reviews, and evaluation. His perspectives on AI development workflows have made his blog and talks a key resource for developers and startups. He has also commented on the limitations of certain AI techniques, notably calling reinforcement learning "terrible" while acknowledging that alternative approaches have their own significant challenges.

Joining Anthropic and Current Focus

On May 19, 2026, Andrej Karpathy announced via a statement on X.com that he had joined the AI safety and research company Anthropic. According to the company, his role involves leading a team focused on research in pretraining, a fundamental stage in developing large language models. This move places him within another leading AI research organization, following his foundational work at OpenAI and his entrepreneurial phase with Eureka Labs. His focus on pretraining aligns with his deep expertise in the core mechanics of how AI models are built and trained from vast datasets, and his current work represents a return to hands-on research at the core of model development.

In his discussions in 2026, Andrej Karpathy has commented on the broader trajectory of AI, suggesting that the development of artificial general intelligence (AGI) is still roughly a decade away. He joined Anthropic in 2026 as part of the pretraining team, where the company stated he will be leading a team for research in pretraining. This role at Anthropic complements his ongoing educational efforts through Eureka Labs, the AI education platform he founded in 2024. Through Eureka Labs, he teaches principles via products like the AI course LLM101n and the broader "Zero to Hero" series on LLM fundamentals. By leading a pretraining research team, he is directly involved in the foundational engineering that determines the capabilities of the next generation of AI systems.

Impact and Public Engagement

Karpathy's impact extends beyond his formal roles through sustained public engagement. His YouTube channel, while starting with Rubik's Cube tutorials, evolved into a venue for sophisticated AI tutorials that break down complex topics. His personal blog and GitHub repositories serve as extensive technical resources for students and practitioners worldwide. This commitment to education was formally recognized when he was described at the Sequoia Ascent 2026 conference as having "become that teacher to the world." His ability to explain detailed concepts in an accessible manner has made him a unique bridge between cutting-edge research and the broader developer community.

His writings and talks frequently analyse the practical and philosophical shifts brought by AI. He discusses concepts like "AI psychosis" in agents and categorizes programmers into new types based on how they integrate AI into their workflow. By openly sharing his own evolving practices, such as how he stopped writing bash scripts manually, he provides a real-time case study of the profession's transformation. This transparency and his foundational contributions to major AI projects solidify his status as a leading voice in understanding both the technical implementation and the societal implications of artificial intelligence.

Questions about Andrej Karpathy

Where did Andrej Karpathy go to university in Canada?

Andrej Karpathy completed his undergraduate studies in Canada at the University of Toronto, where he earned bachelor's degrees in Computer Science and Physics in 2009. He then pursued a master's degree at the University of British Columbia, graduating in 2011. His master's research involved working with physically simulated figures. These Canadian educational foundations preceded his doctoral studies at Stanford University in the United States.

What is Eureka Labs and what does it teach?

Eureka Labs is an AI education company founded by Andrej Karpathy in July 2024. Its primary focus is teaching the fundamentals of large language models (LLMs) through online courses. The company's first product was the course LLM101n. It also offers a broader educational series called "Zero to Hero" on LLM fundamentals. The platform advocates for the use of AI teaching assistants, though this approach has sparked debate regarding data privacy and the student-teacher relationship.

What did Andrej Karpathy do at Tesla?

Andrej Karpathy joined Tesla in June 2017 as the director of artificial intelligence, reporting directly to CEO Elon Musk. In this role, he led the Autopilot Vision team, which was responsible for developing the computer vision systems for Tesla's vehicles. His work involved applying deep learning and convolutional neural networks to enable cars to perceive and interpret their surroundings using camera data. He held this position until his departure in July 2022 after a sabbatical.

What is 'vibe coding' according to Karpathy?

Andrej Karpathy coined the term "vibe coding" in February 2025. It describes a method of software development where individuals, often hobbyists, use AI tools to construct applications and websites primarily by typing natural language prompts. This approach minimizes traditional manual coding. By 2026, Karpathy contrasted this with more rigorous "agentic engineering," which involves structured specification design, code review, and evaluation rather than just a "prompt and pray" mentality.

Where does Andrej Karpathy work in 2026?

As of May 2026, Andrej Karpathy works at the AI research company Anthropic. He announced his move via a statement on X.com on May 19, 2026. The company stated that he is leading a team focused on research in pretraining, which is a fundamental phase in developing large language models. This role follows his earlier work at OpenAI and Tesla, and his founding of the education company Eureka Labs.

What is Andrej Karpathy's YouTube channel about?

Andrej Karpathy's YouTube channel, badmephisto, began in 2006 and initially gained popularity for its detailed Rubik's Cube solving tutorials, which were used by world champion speedcubers. The channel has garnered over 9 million views. In recent years, he has used it to post educational content on artificial intelligence, including tutorials on how to create artificial neural networks. The channel remains an active part of his public engagement and teaching efforts.

Sources: Wikipedia and current public reporting, checked by the Your Space Hamilton desk. Last updated 2026-10-18.