Accelerating innovation: Y Combinator
summary
Claim: 25% of YC batch with 95% AI-generated code. Details: YC handles 800 startups a year and 80,000 applications, reorganizes to avoid turning partners into data-labelers, and processes 12,000 apps for 3,000 AI school spots. Tools and demos: Gemini 2.5 Pro + agent infra loads JSON and Postgres biographical data to match alumni at an event with one prompt, plus windsurf automations inside nontechnical roles. Evals become the early moat; the internal agent demo works.
AI-written summary · about
chapters (11)
- 0:00Intro
- 1:10Garrys first logo
- 1:42Role of design in the age of AI
- 3:55Where does 5 coding shine
- 5:02Large organizations have advantages too
- 6:10What is the difference between Moes and software
- 7:21What lessons should we learn from this
- 9:44How should we plan for new capabilities
- 11:20How to prompt AI
- 13:04Lessons from Y Combinator
- 17:01The importance of the eval
related talks
47mAI Pitch: Beyond the Hype
15mAX is the only Experience that Matters
02mSteve Yegge, Sourcegraph, ETLS Vegas (25)
35mBeyond the Hype and Hyperbole
18mDispatch from the Future: building an AI-native Company
29mAI-Driven Development: Revolutionizing Productivity, Software Development Life Cycle
47mHow To Build a GenAI-Augmented Software Organization
51mAssessing AI's progress