This is Andrej Karpathy’s keynote from Y Combinator’s AI Startup School in San Francisco, hosted on the official Y Combinator channel. In it he argues that software is changing fundamentally for the third time. Software 1.0 was hand-written code, Software 2.0 was machine learning models trained on data, and Software 3.0 is large language models that you program in plain English.
Karpathy frames LLMs as a new kind of computer and explores what that shift means for the people who build and use software. He discusses how natural-language interfaces change the design of applications, where humans should stay in the loop, and why he thinks the field is still in its early, formative period rather than a finished state.
For a business audience, this is a short and accessible strategic talk rather than a technical tutorial. It comes from someone who has led AI at major companies and helped popularize the very “Software 2.0” framing he is now extending, which makes his read on the direction of the field unusually well grounded.