Anthropic's Generational Run, OpenAI Panics, AI Moats, Meta Loses Lawsuits

Welcome back to the number one show in the world—Fantastic Four assembled, brothers in arms, and Sacks beaming in from Texas after a whirlwind week. Big board today: Anthropic’s heater, OpenAI’s pivot, markets re-rating, social-media lawsuits, and Sacks stepping into PCAST with a monster roster.

Texas has been great, even on three hours of sleep, and there’s a ton to get through.

Anthropic’s moving fast—coding first, then co‑work and agents, even phone-to-desktop control—and enterprise spend followed. What’s your read on their run?

I’ve always rated their products; the coding bet unlocked enterprise budgets and became a platform for documents, slides, and agents. My beef is their push for a D.C. permission regime on chips and models, which tilts the field toward incumbents; product-wise they’re firing, and I’ve stayed out of the Pentagon dispute, full stop.

There’s also chatter that Anthropic’s politics are a recruiting magnet; is that real or theater?

It’s authentic to who they are and it resonates with their team; more broadly we’re seeing left‑right versions of almost everything, including AI tools.

Both OpenAI and Anthropic are excellent but not directly comparable; OpenAI skews consumer subs and is conservative on revenue, Anthropic is API‑heavy and books gross usage. Press narratives miss that normalization still shows OpenAI as the bigger revenue engine while Anthropic gains fast in enterprise.

OpenAI still owns consumer mindshare, but share’s slipping as the pie explodes; Sora’s reportedly shelved and they’re trimming side bets while pushing into enterprise and a PE‑style go‑to‑market. Smart reset or wobble?

Focus wins; I’d double down on consumer where the brand is dominant and let enterprise be a separate muscle. Winning the workplace demands different features and a different cadence.

Do consumers pay or does Apple, Google, Meta make it free and ad‑supported?

People already pay for must‑have services at massive scale; a capable AI that handles travel, inbox, calendar, and finances could be the most valuable personal service yet, with paid tiers plus a rich third‑party ecosystem.

Most users will accept ads, and premium will still reach hundreds of millions; I like enterprise stickiness as an investor, but both models can thrive.

Google will fight hard on consumer because search and AI chat are converging, and they already hold your mail, docs, and calendar, which lowers trust friction for agents; I’m waiting for their OpenAI‑style assistant.

They’re the rare company that can run consumer and enterprise as distinct efforts because the cash flow supports it.

PE’s AI rollup wave claims they can force real adoption. Does owning change management unlock the value?

Yes; pilots often stall and value is trapped in the org. If you own the business and the rollout, you can drive the gains instead of hoping they happen.

Markets are splitting between a superintelligence future and the next‑gen software view; we priced like the former, but public multiples are resetting. SaaS got re‑rated to cash-flow math, while Apple, Microsoft, Meta, and Alphabet get durability premiums; if disruption accelerates, employees will push for cash over long‑dated equity.

Winners won’t sit still; the incumbents that embed AI and ship real productivity will take share, not just in software but across complex, asset‑heavy industries.

What’s the moat now—brands, networks, leadership?

I don’t count management as a moat; network effects, hard‑to‑replicate physical capability, and subtle switching costs still matter in a world of digital abundance.

Brand premiums erode when a cheaper, better product shows up; look at Tesla’s impact on luxury autos. Abundance beats old‑school brand equity, except maybe at the ultra‑luxury fringe.

Agents are starting to bypass apps; you ask and it does the work across your tools.

Enterprises are literally asking us to hide the complexity and let them state intent; they want the system to orchestrate payments, travel, and workflows behind the scenes.

You’ll still want a clean dashboard to verify and visualize, even if voice and chat handle most actions; Apple’s role depends on whether agents augment or replace the app wall.

Programming note: Liquidity is now part of All‑In, and our resident ‘melanin farmer’ seized the booking wheel.

We’ve locked Dan Loeb, Sarah Friar, Bill Ackman, and Andrej Karpathy; Andrej will do a quick AI future set, then a fireside. Auto Research and vibe‑coding are letting tiny teams ship in days what used to take months—I’ve seen CTRs double overnight.

I finally built a fifteen‑year‑old idea in a weekend with vibe‑coding; it feels like a business replicator just spun it up.

Two verdicts hit Meta in two days—child‑safety and product‑design harms—reviving the Big Tobacco playbook. Where do you land on liability versus choice?

Tort litigation is a massive tax on the economy, and we underweight personal responsibility; parents should keep kids off social media and screens until they’re ready, and companies should inform clearly, but not every bad outcome is a lawsuit.

When firms know about severe harms and press ahead, that’s different; age checks and clear labels are the baseline we used in alcohol, cars, and hazardous materials.

Informing parents is critical, but age gating is thorny, and the litigation machine will chase deep pockets regardless.

Social use isn’t smoking; harms and benefits are more subjective, so empower parents with controls instead of blanket bans. Be careful turning playlists and feeds into the next Big Tobacco; the LA case had serious causation gaps.

Trial lawyers found a path around Section 230 using product‑liability framing, so expect a flood of cases; I want real parental kill switches and workable age verification, but schools pushing YouTube make it a whack‑a‑mole.

The new AI framework supports age assurance and parental controls; for AI apps especially, parents should decide, because the upside for learning is huge.

On to PCAST: President Trump tapped you to co‑chair; what changes with this remit?

I’m co‑chairing with Michael Kratsios and staying on AI, now alongside nuclear, quantum, advanced semis, and biotech; we invited builders who’ve shipped the core tech we’ll be advising on.

The lineup—Andreessen, Sergey, Dell, Ellison, Friedberg, Jensen, Lisa Su, Zuckerberg—leans operator‑heavy; why that mix?

We also have decorated scientists, and adding doers is a feature; if you’re setting semiconductor strategy, having the people who invented keystone products in the room helps.

I’m honored; this echoes FDR’s original intent but in an AI era, and the China race is real—on publications they’ve surged past us and they’re industrializing discoveries quickly. That’s why industrial scientists and technologists matter right now.

When the President’s Council on podcasting finally appears, we’ll bring our mics to D.C.; until then, that’s another heater from the number one podcast in the world—bye‑bye, love you boys.