How AI is increasing the surface area for serendipity
summary
Claims that AI already writes large swathes of code—41% of GitHub code is AI generated—and that LLMs can pass programmer exams, with sub‑200 ms latency and API glue enabling cross‑team workflows. Examples: a bedtime‑story app for kids built and running, image models generating 800 bananas in seconds, and a batch experiment of 10,000 inflatable chairs that the model scores for selection. Hardware gains like Grock’s LPU boosting inference 10–200× turn these into practical, surprising workflows.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 6:30Human role in AI age
- 9:50AI for creativity
- 13:13AI growth
- 16:42AI and serendipity
- 17:03Speed in AI
- 19:34Generative AI beyond LLMs
- 23:10Conclusion
related talks
27mSoftware engineering with GenAI
30mThe AI Revolution in Coding
46mThe AI-Powered Developer
26mThe present and future of programming with AI (Keynote)
39mAI and the Democratization of Software Development
49mMagic of Automation and Everyday Chores
15mUsing AI effectively in large Java codebases
01h 22mSoftware Abstraction in the Age of AI-Generated Code