Recursive Self-Improvement and the Future of AI

Name: Recursive Self-Improvement and the Future of AI
Uploaded: 2026-10-12T00:00:00.000Z
Duration: 1926 s
Channel: a16z
Description: Martin Casado interviews MakerMaker.AI's founders on recursive self-improvement: what RSI means at prompts, harness, weights, and chips; why their bets start from grounded math; and why simple closed loops that survive scale beat clever research ideas. A substantive take on the field's loudest term.

Aa16z

Full transcript

English

Summary：Martin Casado interviews MakerMaker.AI's founders on recursive self-improvement: what RSI means at prompts, harness, weights, and chips; why their bets start from grounded math; and why simple closed loops that survive scale beat clever research ideas. A substantive take on the field's loudest term.

Watch on YouTube

Core points (3)

1RSI spans prompts, harness, weights, and chips — each layer improves on different timescales
2Simple closed loops beat clever research ideas once tested at economic scale
3Real RSI evidence is capability and efficiency improving together, not demos