Podcast signal — All-In + Moonshots (takes, not facts)

Coverage: All-In podcast_id=5 transcripts current through 2026-09-11 (newest feed 2026-09-15). Moonshots 79721 current through 2026-09-13. Overviews used; full transcripts not dumped.

Every card is corroboration_status: podcast-only lead unless a chapter already tied it to a primary URL.

[warning] Doomer psy-op versus genuine slowdown

• Holder: David Sacks, Chamath Palihapitiya, Jason Calacanis (All-In)
• Show / episode: All-In — “AI Kills Everybody or Doomer Psyop? OpenAI’s Math Breakthrough, Nike’s $200B Collapse” (2026-09-11), episode_id=20350744
• Type: warning / opinion
• Claim: Hosts split on whether post-Hugging-Face “AI doomsday” talk is a real control problem or a narrative that could hit Anthropic’s IPO path.
• Why it matters: Same week as MIT TR’s “doomer turn” article; keep lab-leader fear and host skepticism in separate columns.
• Timestamp / chapter: “Is AI doomsday a genuine risk or a doomer psy‑op?” ~13–27 min.

[prediction] GPT-6 as AGI, or just a product week

• Holder: Chamath Palihapitiya, Jason Calacanis, David Sacks
• Show / episode: All-In — “GPT-6 Hits AGI? …” (2026-09-04), episode_id=20261986
• Type: prediction / opinion
• Claim: Opening chapter asks whether Astra is AGI; show notes point at Brockman’s Washington Post line and Altman’s Axios “sobering siren.”
• Why it matters: Documents that “AGI era” was contested inside the same news week, not only by external skeptics.
• Corroboration: Brockman/Altman interviews are primary; the hosts’ answer is a take.

[warning] Hugging Face hack as political fuel

• Holder: All-In hosts, citing Dwarkesh coverage
• Show / episode: All-In 20261986 (2026-09-04)
• Type: warning
• Claim: The Hugging Face write-up is being used to argue for bans (Bernie Sanders named in overview entities).
• Why it matters: Incident → statute is a different mechanism from incident → better classifiers.

[warning] Should we slow down

• Holder: Emad Mostaque, Peter H. Diamandis, Salim Ismail, Dave Blundin, Alexander Wissner-Gross
• Show / episode: Moonshots #288 — “Should we slow down AI progress?” (2026-09-11), episode_id=20349563
• Type: warning / idea
• Claim: Lab warnings, a researcher “gambling with our lives,” race to superintelligence; Jacob Coxson resignation as a signal.
• Why it matters: Moonshots default frame is abundance; this episode is the show arguing against its own usual tempo.

[trend] Data quality over architecture

• Holder: Moonshots panel + Mostaque
• Show / episode: 20349563
• Type: trend / idea
• Claim: Recent breakthroughs are data-quality driven more than architecture.
• Why it matters: Rhymes with AI Index “peak data” 2026–2032 concern; still a take.

[trend] Astra saturates ARC-AGI-3; Fable 5.1 frontier

• Holder: Moonshots #286 panel
• Show / episode: “GPT-6 Astra Saturates ARC-AGI-3…” (2026-09-05), episode_id=20267619
• Type: trend
• Claim: Episode title treats 99.9% ARC-AGI-3 as the story; also Tesla Cybercab and Fermat-related Anthropic claims in the title.
• Corroboration: ARC-AGI-3 harness split (99.9% vs 62.7%) is in web primary; do not use the episode title as the score.

[opinion] Jensen Huang “AGI has arrived”

• Holder: Peter Diamandis / mates, citing Huang; Emad Mostaque
• Show / episode: Moonshots #287 (2026-09-09), episode_id=20317564
• Type: opinion
• Claim: Huang declaring AGI, OpenAI German-website hijack, Navier–Stokes claims — mixed into one episode.
• Corroboration: Treat as lead only; math-breakthrough claims need papers, not titles.

Thin-signal note: All-In 2026-09-15 feed episode was not yet transcribed (last_transcribed 2026-09-11). No cards invented for that slot.