# Andrej Karpathy Andrej Karpathy (born October 23, 1986) is a Slovak-Canadian computer scientist, AI researcher, and educator. Born in Bratislava, he moved to Toronto at age 15. He earned degrees from University of Toronto (BS Computer Science and Physics, 2009), University of British Columbia (MS, 2011), and Stanford University (PhD, 2015) under Fei-Fei Li. Karpathy is a founding member of OpenAI (2015-2017, briefly returned 2023-2024), served as Senior Director of AI at Tesla leading the Autopilot computer vision team (2017-2022), and founded Eureka Labs, an AI+Education company in 2024. He created Stanford's influential CS 231n course and is known for his educational AI content on YouTube. Named MIT Technology Review "Innovators Under 35" (2020) and TIME100 Most Influential People in AI (2024). ## Frameworks and Coinages Karpathy is the source of much of the working vocabulary for the LLM era. Notable framings include: - The Software paradigm trilogy: [[Software 1.0]] (explicit code), [[Software 2.0]] (learned weights), [[Software 3.0]] (LLM-as-runtime, prompt-as-program). - [[Vibe Coding]] (Feb 2025); the floor-raising mode of LLM-driven coding. He later contrasted it with [[Agentic Engineering]] (the ceiling-raising mode) at Sequoia AI Ascent 2026. - [[AI Verifiability]] and [[AI Verifiability as a Capability Ceiling]] (the four-factor formula: verifiability × training attention × data coverage × economic value) explaining LLM jaggedness. - The "animals vs ghosts" framing for what LLMs *are*: not evolved biological intelligences, but optimization-pressure-shaped statistical entities. - The [[Menugen Architecture Pattern]] and [[MenuGen Deployment Gap]] as worked examples of LLM-native apps and their remaining friction. - [[Markdown-based Installation (MD Scripts)]] as the LLM-era replacement for shell scripts. - [[LLM Wiki]] and [[LLM Knowledge Bases Over Unstructured Data]] as the genuinely-new computational primitive LLMs unlock. - The [[Agent-Native Product Decomposition]] (sensors / logic / actuators) framing for the agent-native economy, with neural computing as the speculative end state. ## Sources to Track - Personal blog: https://karpathy.bearblog.dev/ - Sequoia AI Ascent 2026 talk (April 2026); the canonical source for Software 3.0 and the agent-native economy. - "Verifiability" essay; the foundation for the capability-ceiling argument. - 2025 LLM Year in Review; the source for ghost intelligence and ambient programming framings. ## Quotes <!-- QueryToSerialize: LIST FROM #type/quote AND [[Andrej Karpathy]] WHERE public_note = true SORT file.name ASC --> <!-- SerializedQuery: LIST FROM #type/quote AND [[Andrej Karpathy]] WHERE public_note = true SORT file.name ASC --> - [[Tools to enhance understanding]] - [[You can outsource your thinking, but you can't outsource your understanding]] <!-- SerializedQuery END --> ## Books <!-- QueryToSerialize: LIST FROM #books AND [[Andrej Karpathy]] WHERE public_note = true SORT file.name ASC -->