First screen, captured 5 hours ago. Open the video.

The avalanche that hit math is coming for AI development

What HN thinks

Half the thread is drafting plans for the end of humanity, the other half is checking who sponsored the podcast.

Insiders say acceleration is real
Some practitioners, including a research engineer at an Alphabet company, report agents rewriting models and running experiments faster than humans can guide them. They feel their own obsolescence approaching.
Hype, bubble, or talking his book
Skeptics argue that math's verifiability doesn't carry over to messy domains, and that agents still need heavy harnessing. They add that AI output hasn't improved products and that the labs' behaviour (chasing money, IPOs) reveals they don't believe in imminent ASI.
Power concentration worries
Others focus on who controls AI, fearing an oligarchy or existential risk. They debate pausing development, open models, and whether raw materials would matter more than money.

Consensus: There is no consensus. The thread splits between people who think recursive self-improvement is underway and people who see an S-curve and influencer hype.

Recommended in the comments

Superintelligence: Paths, Dangers, Strategies (book): Recommended as the book lab leaders have read and as background on how an ASI could win decisively.
Lean (tool): Cited as part of a harness that gives research agents a strict sense of reality.
SageMath (tool): Mentioned as a tool that grounds agents doing math and CS research.

Summarised by Claude from the comment thread 4 hours ago.

Points: 44
Comments: 88
Best rank: #4
Posted: 2026-11-13 by crescit_eundo
Category: AI & ML

superintelligence recursive-self-improvement ai-research-automation agi-timelines ai-mathematics dwarkesh-patel

All captures:

Full page

Captured 5 hours ago. Click for the whole screenshot. The page is longer than 25,000px, so the screenshot stops there: read the rest on dwarkesh.com ↗

Looks similar

Screenshot of planned-obsolescence.org: AI is dramatically accelerating AI research

AI is dramatically accelerating AI research (planned-obsolescence.org)

333 points by yurivish 7 days ago | 300 comments

Ajeya Cotra's Substack post arguing a METR report shows AI is already roughly doubling AI capabilities progress at Anthropic, implying rapid compounding toward superintelligence.

ai-safety recursive-self-improvement anthropic metr