Kevin Roose is AGI-pilled. Why aren't you?

What the episode covered

In this partial transcript, Kevin Roose said LLMs are the leading path to AGI and that AGI belief helps frontier labs make large compute bets, while Nilay Patel said current systems still feel brittle. The register is mixed, with friendly podcast banter, careful uncertainty, and several AI-adjacent sponsor reads surrounding the discussion. Watch whether labs show recursive self-improvement, whether coding gains generalize beyond verifiable tasks, and whether compute advantage remains decisive.

Our summary, written from the transcript. Not the speakers' words.

The episode, in order

1. 01Machine Gods and leaving The New York Times

   The launch gives Roose a direct business stake in the media format question he is discussing.

2. 02YouTube as a media platform

   This explains why the media strategy discussion is not just promotional chatter, but part of Roose's broader view of where journalism can grow.

3. 03The book as historical record

   That positions his reporting as preservation of source material, not only industry analysis.

4. 04The AGI shift

   This is the central technical and cultural claim in the conversation.

5. 05Enterprise software and AGI ambition

   The exchange separates public revenue mechanics from the more speculative end-state beliefs inside AI labs.

6. 06Patel's AGI skepticism

   That gives the episode a concrete counterweight to Roose's stronger belief in the LLM path.

Sponsors named in this episode

No sponsor named doesn't mean no ads. Each row was taken from the ad break as it was read on air. The verbatim passage behind every row is checked against the transcript character for character. Short excerpts appear below.

Roles and stated affiliations

Sponsor or conflict flagged. AI-adjacent sponsor adjacency — The recording includes ads for CrowdStrike AI-DR, AWS AI, OutShift multi-agent AI, Framer AI agents, and Shopify Sidekick during a conversation about AI business and AGI. No disclosure was stated on the recording that tied any sponsor to the guest, the book, or Machine Gods.

Wakewire's read of interests

• Kevin Roose: He is presented as the author of The AGI Chronicles and co-founder of Machine Gods, so the episode discusses both his reporting claims and his new independent media business.
• Nilay Patel: He hosts Decoder as editor-in-chief of The Verge and frames the interview around media strategy, AI company structure, and his own skepticism about AGI.
• CrowdStrike: The commercial copy promotes AI Detection and Response and directs listeners to CrowdStrike.com slash decoder.
• AWS AI: The commercial copy promotes AWS AI as a way to use multiple models, agents, speed, and security.
• Named advertisers: Amex Corporate, Quo, Framer, Shopify, and OutShift are presented through commercial copy promoting business services, website tools, commerce software, or AI resources.

Recording limits

• The transcript cuts off mid-sentence at the end, so the final exchange about 'software brain' is incomplete.
• Several passages appear to have missing or corrupted words, including 'long supalm', 'Dario Amadei Supra belief', and 'Support factual claim'.

Claims that need more support

• Kevin Roose said OpenAI is the only major AI company that is not severely compute constrained, but the episode did not provide supporting capacity figures or a source.

Claims made on air that the episode itself did not back up.

Are you from this show? You can ask for a correction, reply, or ask us to stop coverage on the rights page.

What was said

• Machine Gods and leaving The New York Times

  Roose said he and Casey Newton left to start a new tech podcast that is intended to be video first and distributed across YouTube, Apple Podcasts, Spotify, and other platforms.

• YouTube as a media platform

  Roose said YouTube is the most important media platform and that high-quality journalism is still a small part of its ecosystem.

• The book as historical record

  Roose said AI history may not leave a durable paper trail because many builders are private-company employees, messages auto-delete, and researchers are too busy to keep notes.

• The AGI shift

  Roose said AGI moved from an old idea without a working technical path to a live race once transformers and large language models appeared to scale.

• Enterprise software and AGI ambition

  Roose said B2B SaaS is the game AI labs have to play to finance compute and research, while Patel pressed whether these companies can operate at that scale.

• Patel's AGI skepticism

  Patel said he is not AGI-pilled because current AI systems still feel brittle when tasks lack clear verification.

Our paraphrase, not the speakers' words.

The passages we relied on

Short excerpts, credited to the show, at most 40 words each and 150 per episode. We do not publish transcripts. We link to the episode so you can hear it from the show itself.

What to watch for

• A frontier lab publicly demonstrates recursive self-improvement in coding models, with the model improving the system that trains or deploys it rather than only scoring higher on coding benchmarks. (frontier model releases)
• OpenAI, Anthropic, or Google describes enterprise AI revenue or committed enterprise usage as directly funding new compute buildouts. (company updates)
• An agentic AI product completes messy, non-verifiable workflows such as government forms with published error rates and recovery behavior, not just a demo. (product evaluations)
• Open-weight models either keep pace with closed frontier models or fall behind after labs restrict distillation. (model release cycles)