Skip to content
Y CombinatorY Combinator

Paul Buchheit: Why Evals, Not Code, Are the Real AI Moat

Predicting the next token dissolved the paperclip maximizer fear; now eval sets, not codebases, are the moat, as Jerry shows with 50% growth post-GPT-4.

Garry TanhostPaul BuchheitguestHarj TaggarhostJared FriedmanhostDiana Huhost
Jan 24, 202539mWatch on YouTube ↗

CHAPTERS

  1. 0:00 – 1:17

    Intro

  2. 1:17 – 5:04

    Retreat.

  3. 5:04 – 8:00

    Demand for AI.

  4. 8:00 – 10:00

    Evals.

  5. 10:00 – 14:37

    Product iteration.

  6. 14:37 – 16:16

    Balance.

  7. 16:16 – 19:31

    Automation.

  8. 19:31 – 22:21

    Predictive models.

  9. 22:21 – 24:58

    OpenAI.

  10. 24:58 – 26:23

    AI tools.

  11. 26:23 – 29:30

    Cursor.

  12. 29:30 – 33:16

    Scaling.

  13. 33:16 – 36:37

    ROI.

  14. 36:37 – 38:30

    Startup success.

  15. 38:30 – 39:32

    Outro

Get more out of YouTube videos.

High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.

Add to Chrome