E290 · All-In Alpha

conviction thresholds crossed by this episode — 65 / 45 / 15 · ideas born here show where they landed. why 65 isn't always green

▲ +43.2 🤖 Meta becomes America's open-source AI champion ember → green threshold 27.5 → 70.7 still watch — needs 2 besties on board

▲ +46.7 🤖 Personal AI agents put the App Store's 30% take and opaque subscription/retail economics on notice born at watch 46.7

▼ -33.5 🌍 The 2026 US boom — 5-6% GDP prints watch → ember 59.1 → 25.6

▲ +18.0 🤖 Frontier AI labs keep durable premium pricing (not commoditized) dormant → ember 0.0 → 18.0

▼ -9.6 📈 2026 is the year of mega-IPOs ember → dormant 16.9 → 7.3

each voice's push on conviction this episode, in conviction points — supports and opposes · share = % of this episode's movement

Chamath

8 captures · 39% of movement · 1 idea born

+58.2 / -38.4 → net +19.8

Jason

11 captures · 25% of movement

+42.2 / -19.2 → net +23.0

Friedberg

8 captures · 19% of movement

+46.1 → net +46.1

Sacks

6 captures · 17% of movement

+33.9 / -7.9 → net +26.0

NEW ▼ AAPL 🤖 Personal AI agents put the App Store's 30% take and opaque subscription/retail economics on notice watch 49 POD DISAGREES ▲ +46.7 0.0 → 46.7

Consumer agents like Meta's Muse and xAI's Grokbot transact headlessly on the user's behalf, so services no longer need an app-store front door and price discovery strips out the breakage that opaque subscriptions and retail rely on. Pressure on Apple's App Store take rate and on companies that profit from opacity, while agent-native rails (Shopify's store API, Stripe) and the agent platforms (Meta) gain.

plays AAPL ·primary AMZN META ▲ NYT SHOP ▲ evals 2027-09-26

Chamath support ×2 explicit_prediction ▶ 1:13:53

The second thing though is really interesting. Things like Grokbot and Muse really put the App Store and its 30 percent rev share on notice. Why? Because all of a sudden, what you do is you force many of these services to exist headlessly, right?

Jason support ×1 sentiment ▶ 1:16:56

Shopify, Sacks, added API access to all the Shopify stores there. So this seems like maybe Amazon's making a strategic mistake here by not embracing the tech.

▲ META 🤖 Meta becomes America's open-source AI champion watch 55 POD DISAGREES ▲ +43.2 27.5 → 70.7

Chamath support ×2 sentiment ▶ 5:25

Zuck last week said, we were about to roll out Muse, which looks like one of the most successful product launches in recent memory, has huge implications actually, which we should talk about.

Friedberg support ×1 sentiment ▶ 13:00

They took their time doing this Muse release this week. They said they wanted to get it right. They wanted to make sure that it was safe, that it didn't give people capabilities to do bad things. That's the responsibility that they took as a corporation.

Jason support ×3 explicit_prediction ▶ 1:07:58

But man, agents as expressed through Muse feels like the first time tens of millions of Americans are going to get some demonstrable value from AI, you know, things that will make their lives better. Muse hit number one last Friday in the App Store. Meta stock was up 10% after this new agent, Muse, was released.

Sacks support ×2 sentiment ▶ 1:11:07

Mark Zuckerberg said in his essay that he thought that AI capabilities should be de-centralized and here he is walking the walk and creating a product that's very easy to use.

▲ RSP 🌍 The 2026 US boom — 5-6% GDP prints watch 65 ▼ -33.5 59.1 → 25.6

Friedberg support ×1 sentiment ▶ 20:03

Maybe we are seeing the AI wave under way right now, the productivity wave, the productivity boom.

Chamath reversal ×2 sentiment ▶ 1:21:55

If you think that you have 3.5% inflation and 5% nominal GDP growth, what you have is 1.5% real. And the problem with that is that AI is probably the majority, if not all of it.

▲ AAPL 🤖 Apple: the AI dark horse via local models watch 47 POD DISAGREES ▲ +21.9 21.4 → 43.3

Friedberg support ×1 sentiment ▶ 20:03

You can have an NVIDIA chip or have a good Mac Studio computer, and you can run this insane model on your desktop computer for free.

Jason support ×1 sentiment ▶ 24:36

Have you loaded these models on your computer? Because I'm playing with it on my MacBook Pro and it's awesome.

▲ MSFT 🤖 Frontier AI labs keep durable premium pricing (not commoditized) ember 38 POD DISAGREES ▲ +18.0 0.0 → 18.0

Chamath oppose ×3 explicit_prediction ▶ 26:10

It'll force OpenAI and Anth to go up the stack. There is no choice. You can't sit there and serve a token that has diminishing value because then all the subsequent value is getting absorbed and made by people wrapping your intelligence token.

Friedberg oppose ×2 explicit_prediction ▶ 41:15

there's a stat that I think in the last 12 weeks alone, token use has flipped from 80-20 closed versus open, to 80-20 open versus closed.

Jason oppose ×2 sentiment ▶ 44:31

the vertical language models, the small language models are going to do 80 or 90% of the tasks that many corporations need

Sacks support ×3 explicit_prediction ▶ 45:21

I think that Anthropic and OpenAI right now are a stable duopoly for frontier intelligence because they have a pretty significant lead over all these other companies, and they are able to charge a premium for that.

▲ NBIS 🤖 Enterprise AI sovereignty: workloads leave frontier APIs for private stacks green 94 ▲ +12.2 74.6 → 86.7

Chamath support ×2 explicit_prediction ▶ 26:10

Now, you may choose to go down to cheaper and cheaper Anth models, so then Anth keeps the revenue, but there's also a risk that you may do what Friedberg just said and just go to an open source model and just host it yourself.

Friedberg support ×3 explicit_prediction ▶ 38:08

Separately, I think most of the enterprise applications are going to be run using these open weight models.

Jason support ×2 sentiment ▶ 44:31

We need people to use those models so they have sovereignty, so you're not giving your data to the frontier model so that they don't get that tip of the spear superintelligence edge, which is why people are disengaging.

Sacks support ×1 sentiment ▶ 45:21

I fully understand that the majority of tokens will go to open models or commodity intelligence.

▲ NVDA 🏛️ The AI doomer campaign fails — no FDAI, open source survives ember 31 POD DISAGREES ▲ +11.7 74.6 → 86.3

Sacks support ×2 sentiment POLICY-ADJACENT ▶ 1:34

And the president called in and did what he did best, which is seize total control of the narrative that I'm sure these Dumer groups have been carefully planning for weeks and kind of thwarted that whole operation.

Jason support ×1 sentiment ▶ 3:15

the nerds were losing their mind last week, and the adults came, daddy came, Elon came, Jensen came, all the adults in the room came, and they just said, listen, if you have software that's defective and not ready to put in the world, maybe don't put it in the world

Friedberg support ×3 explicit_prediction ▶ 20:03

We are past the point of Bernie Sanders, put the super intelligence away, it can never be allowed in the real world. This stuff is out. The open-source, open-weight models are free. They are on the Internet. You can go and download them, and you can run them on your computer at home.

Chamath support ×1 sentiment ▶ 1:04:49

I'm saying if you stop and you hit the pause button through some regulatory mechanism, you effectively allow six companies to over-earn.

▲ IPO 📈 2026 is the year of mega-IPOs ember 43 POD DISAGREES ▼ -9.6 16.9 → 7.3

Jason oppose ×1 sentiment ▶ 29:19

And their IPOs are looking like they're both going to get delayed. Anthropic is just targeting a $2 trillion valuation, OpenAI at $1.2 trillion. Sam already said OpenAI plans on doing 2027 because of safety concerns.

▲ GOOGL 🤖 Anthropic IPO (~Oct 2026) is attractively priced at $1.5-2T watch 61 POD DISAGREES ▼ -7.9 7.9 → 0.0

Sacks oppose ×3 sentiment ▶ 30:44

So think about the risk factors that's creating. How do you IPO with your own management of the company saying that?

Jason oppose ×2 sentiment ▶ 31:54

We would put the CEO in the lab, and we'd find a CEO who knows how to run the company and not sabotage it. He's literally sabotaging his own company.

Chamath oppose ×3 explicit_prediction ▶ 34:21

Even if you think that they're on a hundred billion dollar run rate, what would otherwise in a very clean sheet IPO maybe would have been a two trillion dollar market cap? Now, you're at one or less.

Friedberg oppose ×2 sentiment ▶ 41:15

That is the biggest risk factor for Anthropic, is they were first out the gate to get the enterprise on AI, to build tools and whatnot. But as this market shifts to open source and open weights, boom, here it is, right here.

▲ NVDA 🤖 Nvidia + Hugging Face makes NVDA the open-source AI champion watch 55 ▲ +5.2 56.5 → 61.7

Jason support ×1 sentiment ▶ 48:14

You have very significant competition, not just from the Chinese, but you have Jensen and NVIDIA, you have also Microsoft, Satya, Google, and Muse at Zuckerberg's house. They're all embracing open-source models and they all have hosting now.

▲ PANW 📈 AI cyber: the one-time hardening upgrade cycle green 78 ▲ +4.0 66.4 → 70.4

Jason support ×1 sentiment ▶ 15:48

To build a product, to do continuous cyber defense. And also a friend of the pod, George Kirch, from CrowdStrike. He also released Falcon, his new product, to do cyber defense.

▲ NVDA 🤖 AI compute capex supercycle has 12-24 months of runway green 86 POD DISAGREES ▲ +3.0 88.2 → 91.1

Sacks support ×2 sentiment ▶ 45:21

To the extent that compute gets scarce, and we're in a compute crunch, just the fact that they've made these huge investments, they're building out so much capacity, and they have those economies of scale.

Chamath support ×1 sentiment ▶ 1:21:55

No. I think that Sacks is probably right on the margins. I think the bigger point is what he said before is, man, the entire economy is effectively levered to this AI trade right now.

▲ GOOGL 🤖 Personal AI agents: Google's data moat wins the form factor ember 23 POD DISAGREES ▼ +0.0 29.2 → 29.3

Jason oppose ×1 sentiment ▶ 1:10:16

wow, how did Google miss this? Man, Google has got to get in the game, but I understand they have this product ready to go.

Friedberg support ×1 sentiment ▶ 1:20:59

I would only feel comfortable connecting all my Gmail and personal Google stuff to a Google service. I don't want to hand over all of my email to someone else.

written during extraction and stored in data/extractions/ep290.json — the auditable source of truth, including everything market-adjacent that did not earn a capture

E290 (post-Summit, all four hosts, Sacks calling in) is one long AI-economics debate with three market-relevant threads. (1) Open source is eating the token market: Friedberg's 10-day release roll-call (DeepSeek 4.1 Flash, Qwen 2.1, Xiaomi MiMo on par with Opus 5, Bonsai 2 on a Mac Studio) plus the viral 12-week flip from 80/20 closed-to-open to 80/20 open-to-closed; Chamath says models are converging within margin of error, the token-selling trade is going away and OpenAI/Anthropic will be forced up the stack into cyber, law and support; Jason adds Anthropic and OpenAI both cut token prices 50% and Jane Street signed $19B of CoreWeave/Crusoe capacity to run open models. That reinforces enterprise-ai-sovereignty (4 voices) and is a 3-on-1 oppose of frontier-ai-premium, with Sacks the lone supporter ('wearing my hat as an investor': stable Anthropic/OpenAI duopoly, 10-30% of the market pays a huge premium, revenue keeps growing while token share falls, but 'the moment they stop being frontier, they go to zero'). (2) The Anthropic IPO is trashed 4-0: Chamath keeps Dario as CEO but says a clean-sheet $2T print on a $100B run rate is now 'one or less' and it clears 'at a much lower price than anybody thinks' (explicit, strength 3); Sacks catalogues the S1 risk factors (leadership's >10% extinction odds, 'pace the frontier' essay days before Opus 5.5, a wet lab after a bio-risk essay, super-voting-share fight, open source), 'they need a psychiatrist, not a banker'; Jason would replace the CEO; Friedberg says the biggest risk factor is customer concentration in fungible tokens. Jason also reports both IPOs slipping (OpenAI to 2027, Anthropic to November or later, Polymarket 96%→76%), logged as a strength-1 oppose on ipo-boom-2026. (3) Meta's Muse: #1 in the App Store, 3M downloads in 10 days, META +10% on release, Chamath on the TestFlight and calling it 'one of the most successful product launches in recent memory'; Sacks ties it to Zuck's decentralisation essay ('walking the walk'). Attached to meta-open-source-ai-leader at thesis level ('free agents for everyone'), even though Friedberg listed Muse among the closed releases: the idea is Meta-as-AI-hero-for-everyone and META is the play. NEW IDEA ai-agents-disintermediate-app-store (Chamath, 1:13:53-1:16:47): agents make services headless, 'really put the App Store and its 30 percent rev share on notice', price discovery kills opacity-dependent businesses (NYT/WSJ roach motels, Amazon blocking bots vs Shopify opening its store API). Bearish AAPL primary; SHOP and META adjacent carry an explicit bullish direction; AMZN and NYT adjacent inherit bearish. Also reinforced: ai-doomer-regulation-fails (Friedberg strength 3: 'we are past the point of Bernie Sanders', a ban is unenforceable against downloaded open weights; Sacks on the Summit thwarting the doomer psy-op and the administration refusing any liability waiver, sacks_policy true; Chamath and Jason at strength 1), ai-compute-capex-supercycle (Sacks: compute crunch, 60% of new worldwide compute is for the two labs, WSJ 'biggest buildout in US history'; Chamath's 'No' to the Oracle force majeure being the top of the AI trade, hedged with 'tread carefully'), apple-ai-dark-horse (Friedberg and Jason both running open models on Mac hardware, strength 1 each), nvidia-open-source-champion and ai-cyber-upgrade-cycle (Jason, strength 1 each). us-boom-2026: Chamath's macro read is 3.5% inflation, 5% nominal, 1.5% real and 'AI is probably the majority, if not all of it', logged as REVERSAL (orchestrator check 2026-09-27: his last stance on the idea was support at E277, 'the market's going to the moon', after an earlier E269 oppose and E272 re-support, so this is a legitimate second flip under the once-per-support rule). Friedberg's 'maybe we are seeing the productivity boom' is a strength-1 support. ai-personal-agents-google split by leg: Friedberg would only hand his Gmail to a Google agent (support 1, the data-moat leg); Jason 'how did Google miss this? Google has got to get in the game' (oppose 1, moderate confidence, he adds Google has a product and a frontier model ready). The Sacks/Chamath Muse enthusiasm was NOT attached here because it credits Meta, not Google. Possible-reversal flags for the orchestrator: Chamath oppose on frontier-ai-premium and on anthropic-ipo (he says the token-premium trade 'is going away' as if he had held it; if the DB shows prior Chamath support, rescore will downgrade to oppose anyway). Note Chamath is bullish the Anthropic BUSINESS (cannot win a recruiting bake-off against them) while bearish the IPO price. LABEL DEFECTS: 1:03 label Chamath 'I'm reliving my childhood, Chamath, with no trauma' is Friedberg's line (continues his Universal Studios remark), non-capturable. SPEAKER_5 13:38 = Jason ('let's move on'); SPEAKER_6 19:11 'Ohalo is a corporation, this is the All-In lab' is a host joke (Friedberg or Jason), non-capturable; SPEAKER_7 19:22 = Jason. CLIPS: SPEAKER_6 57:02 = Trump UN speech, SPEAKER_5 58:19 = Bessent, SPEAKER_7 58:58 = Obama, all cued by Jason and none captured; Sacks 57:31 paraphrases the Trump clip, not clip audio. Remaining SPEAKER_6 turns (53:10, 53:36, 1:03:06, 1:03:08, 1:05:54, 1:26:14, 1:26:58) are one-line host interjections. Minor merges inside turns: 'Hypocritical. Hypocritical.' inside Sacks 7:05, 'Yes.' inside Friedberg 20:03, 'Oh, here he goes' at the end of Friedberg 1:07:21, all Jason interjections, none quoted. All four hosts verified against fingerprints (Chamath: 8090 labs function, 2007 Facebook social ads, recruiting against Anthropic; Jason: intro/outro, Blippi investor, world's greatest moderator; Sacks: investor hat, 'people in Washington', Austin/moving back; Friedberg: runs a life-sciences R&D org, executive producer of the Summit, wet-lab science corner). Every host question was answered on the right label. No time holes (20:03-24:36 is one Friedberg monologue) and no replayed old tape. check_batch WARNs are not expected since no capture was re-attributed. DECLINE LIST: Friedberg 'you don't even need data centers to do 90% of what you can do with AI' (20:03) is an argument that a ban cannot stop AI, not a capex demand call, so not an oppose on the capex supercycle; Chamath '$10T endowed to 3-6 companies' (1:01:32) is conditional on a regulatory pause, he retracts the unconditional version at 1:04:49; Chamath 'labs will be forced to do cyber' (16:28) is lab strategy, not a stance on the PANW/CRWD upgrade cycle; Sacks 'hamster wheel, they go to zero if they slip' and 'FDAI would commoditize them' are conditional, not an independent-frontier-labs-lose call; Jason's Jane Street $19B CoreWeave/Crusoe receipts speak to sovereignty, not to neocloud durability; Jason 'Amazon making a strategic mistake' is folded into the new app-store idea rather than an amazon-compounder oppose (he also says the bot makes him spend more on Amazon); Chamath's 'much more moderate IPO motion' still assumes the IPO happens so it stays on anthropic-ipo, not ipo-boom; Friedberg 'Anthropic has the best life sciences models' is not read as an oppose on google-ai-winner's specialised-model plank; Oracle force majeure over natgas permits is a datapoint, not a stance on natgas-powers-ai or data-center-backlash-reverses (nobody predicted the backlash reverses or persists); Sacks 'Democrats would drive AI offshore like crypto' is conditional on a Dem sweep; Chamath's frontier-labs-up-the-stack and Friedberg's Anthropic-wet-lab-to-therapeutics threads have no listed instrument and were not coined. REGISTRY SWEEP: 12 of 57 ideas hit. 45 positively ruled out as not discussed or discussed without a directional stance, including spacex-multiple-ways-to-win (only as an S1 example), us-debt-spiral-long-rates (3.5% inflation cited as an assumption, no yield call), enterprise-saas-moats, software-industrial-complex-2026, hyperscaler-clouds, tech-melt-up-pre-peak (the 'is this the top' answer is logged on capex), data-center-backlash-reverses, natgas-powers-ai, tesla-second-act, google-ai-winner, independent-frontier-labs-lose, neocloud-arbitrage, amazon-compounder, meta-ai-assistant-takes-search-share (Muse is an agent, not search), everything-becomes-a-market and gambling-speculation (Polymarket only as odds), ai-gaming-entertainment (agent-generated Wordle is not the gaming thesis), section-230 and social-media-tort (Jason's liability-shield rumour is about AI labs), fed-easing-2026, china-manufacturing-automation-edge (Xi's 100k-Americans invite and the shipbuilding-emperor analogy are not the manufacturing thesis); the rest were absent outright.

Anthropic IPO at Risk, Meta’s Muse Pop, Token Prices Fall, Open Source Gains Share, Alignment Fails

Tier crossings

Who moved the board

What got argued (13 ideas)

Episode digest