Pular para o conteúdo

AI Did Brain Surgery on Another AI (Then $1.5T Vanished)

5 de novembro de 2026 · 1 h 8 min

Notas do episódio

FRONTIER

01:24 AI editing AI: Reading and rewriting neural-network weights

03:11 Small models, big math gains: What did the experiment show?

05:06 Weight editing versus fine-tuning

06:03 Could this unlock continual learning?

07:38 Surgical alignment: Fixing specific model behaviors

10:08 Could AI rewrite its own weights while running?

12:41 Are open weights becoming the equivalent of source code?


SECURITY NEWS

13:13 A fake Claude Code extension reaches Cursor users

14:34 App-store impersonation and software supply-chain attacks

16:14 Why wasn’t this caught before users installed it?


ECON / BUSINESS

18:55 Is the AI bubble over? The tech sell-off and fundraising pressure

20:45 Google’s earnings: Cloud growth versus AI spending

21:07 Are AI answers eating Google’s advertising business?

23:34 The innovator’s dilemma: Disrupt yourself or lose

25:45 Microsoft, Amazon, and the AI infrastructure race

27:39 Reddit: Rising revenue, falling shares, disappearing search traffic

28:55 Is AI pushing the open internet behind sign-ups and paywalls?

30:53 What happens when agents bypass the businesses they use?

32:22 Anthropic’s growth and the prospect of trillion-dollar AI companies


ROBOTICS

35:50 Robot imports and tariffs: Security concerns versus supply-chain chaos

38:56 Why trade matters for manufacturing—and peace

40:00 Kroger’s inventory robot and Meta’s humanoid demo

41:08 Marty’s experiments: AI-designed schematics and circuit boards

42:12 From circuit boards to CAD: Vibe coding for hardware

42:37 Building a robotic dog from an AI image


FRONTIER / WORLD MODELS

45:08 JEPA versus LLMs: Do different models learn the same reality?

47:14 World models versus next-token prediction

50:17 Inference, grounding, and whether alternatives can catch LLMs


QUICK NEWS / TECH

52:16 Google’s new frontier-model announcement—and the wait for access

53:58 Can a sleep-inspired approach help AI keep learning?

54:39 Genie 4: Playable worlds generated on the fly

57:23 AI game clones, decompilation, and the copyright debate


SCIENCE

1:00:11 AI-assisted drug discovery and antibody design

1:01:16 AI peer review: When humans prefer the machine’s feedback

1:02:11 Should every research paper get an AI review?


ARTS / CULTURE / WEIRD

1:02:50 An AI Odyssey: Good visuals still need good directing

1:05:24 MC Hammer’s AI future—and his Silicon Valley past

1:06:24 Sam Altman, data centers, and the RAM jokes

1:07:04 A $60 million AI supercomputer on eBay

1:07:31 LGTM: An AI-made takedown of AI coding, plus the sign-off

Each week, join Chris Slowe (founding engineer, former CTO, and now Senior Technical Fellow @ Reddit), Marty Weiner (founding engineer at Pinterest, former CTO at Reddit, and co-founder at VerifyYou), and Zach Weinersmith (creator of SMBC Comics and co-author of NYTimes bestsellers including A City on Mars and Soonish) for the latest in AI and Tech News.


Todos os episódios de Last Humans in the Loop

Configurações

Exibição

Idioma

Define quais estações você vê primeiro na página inicial.

Histórico

As rádios e episódios ouvidos recentemente e suas últimas buscas ficam só neste navegador.

Avisos

Um aviso rápido quando um podcast que você segue tem um episódio novo. Desative podcasts individuais pelo sino na página deles.

Conta

Salve favoritos – no app e no site.

Atalhos de teclado

/
Abrir busca
Espaço
Reproduzir ou pausar
← →
No player: voltar 15 s, avançar 30 s (episódios) · na fileira: próximo bloco
↑ ↓
Na fileira: alvo dentro do bloco
Home / End
Na fileira: primeiro ou último bloco
Esc
Fechar janela ou menu
?
Esta visão geral
Tocando agora

Ir para a estação

A seguir

    Adicione episódios com “A seguir” – eles tocam depois do episódio atual.