Anthropic Wants to Slow AI Down & Instinct Wants a $10B Valuation. Is It Because They're Losing? Is Dario Amodei Actually Right?

Finally, a full house week where Sam is back from New York and the squad is back for a full dive into this week’s AI drama. Anthropic's Dario wrote a long essay calling for the AI race to slow down, weeks before a multi-trillion dollar IPO. Dave's read: Anthropic must be losing. Sam's read: the labs are bad businesses with no compounding advantage, so regulatory capture is the only move left. 

Also this week: Sam's theory that OpenAI and Anthropic only become valuable as defense companies holding unreleased models in reserve, the poly-agent debate over whether normal people want one agent or ten, Instinct raising at $10 billion two weeks after $2.5 billion, and why everyone in the Valley is trading instead of investing.

In this episode:

- Why Dave thinks Dario’s slowdown essay is a sign Anthropic is falling behind, and what OpenRouter's data shows

- Sam on why there's no return on being first, and why that pushes the labs toward regulatory capture

- Brit on what's real in the safety debate, and the damned-if-you-do problem of racing China

- What DeepMind's new boss and Satya Nadella think about building safety alongside the product

- Sam's three eras of tech policy, from Bill Gates to embrace and extend

- The defense company thesis: zero-day models held back like nuclear weapons in silos

- Alignment's next frontier: is your agent working for you, or for the lab that made it?

- Instinct's $10B round, OpenRouter's exit, and why Sam says he's "currently wrong" as a seed investor

Chapters:

00:00 Intro

01:02 Not Dead Yet From the AGI

03:31 Jacob Coxon Quits Anthropic and Goes on a Media Tour

04:47 Dario's Slowdown Essay: Is Anthropic Just Losing?

06:13 Astra, OpenRouter, and DeepSeek Closing the Gap

06:45 Why the AI Labs Have No Compounding Advantage

07:29 Independent Assessors, Bioweapons, and Racing China

09:07 You Don't Need Superintelligence to Break the Banks

10:23 What DeepMind's New Boss and Satya Think About Safety

12:32 OpenAI's $1.5T Raise and the IPO That Keeps Slipping

13:03 Three Eras of Tech Policy: Gates, Google, Embrace and Extend

15:02 What DC Really Thinks About Anthropic

15:42 Should the Labs Red Team Each Other's Models?

18:42 Sam's Take: The AI Labs End Up as Defense Companies

20:54 10,000 Superintelligences in Silos

21:38 Alignment: Does Your Agent Work for You or the Lab?

23:44 The Poly-Agent Future: Muse, Grok, and Claude

24:47 Brit Pushes Back: Normal People Only Want One Agent

26:38 Oura's Agent Explains Jess's Sleep Data

28:40 Instinct Raising at $10B: Are These Numbers Fake?

30:40 Everyone's Trading, Nobody's Investing

33:27 Sam Built a Torah and Talmud App With Claude

35:28 Apple Eyes the Server Business Again

35:58 Sleep Age Wars: 8 Sleep, Oura, and Made-Up Scores

37:22 Entertaining Ourselves to Death

41:01 Building the Things Nobody Else Will Build

43:07 Tech's Democrat Problem: Jay Carney and Ron Conway

48:25 Doctor Strangelove and Not Looking at the Big Board

49:53 Pop Culture: Emmys, Taylor Swift, Silo, and Rogue Heroes

52:15 The WTF Lineup and the One-Name Theory

54:20 The Information's Most Influential People in Tech

We're also on:

X: https://twitter.com/moreorlesspod

Instagram: https://instagram.com/moreorless

Spotify: https://open.spotify.com/show/2CJdt6HhJ8ioNkbqqVKckC

Apple Podcasts: https://podcasts.apple.com/us/podcast/more-or-less-from-the-morins-and-the-lessins/id1692234179

Connect with us here:

1) Sam Lessin: https://x.com/lessin

2) Dave Morin: https://x.com/davemorin

3) Jessica Lessin: https://x.com/Jessicalessin

4) Brit Morin: https://x.com/brit

About More or Less: For 15 years, the Morins and the Lessins have debated the future of Silicon Valley as the closest of friends. Every Friday, they break down technology, capital, and media. From The Information, Offline Ventures, and Slow Ventures.