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

Episode 290 opens not on markets but on the fifth annual All In Summit, and Jason Calacanis immediately turns the floor over to David Friedberg, who produced it. Asked to rate his own performance as executive producer, Friedberg does not rate himself; he spends the answer on a production-team shout-out he says they forgot at the end of day two. The exchange is small talk, but it sets the register for everything that follows: the hosts treat the summit as an institution of their own making, and much of the news they then discuss is news they helped stage.

David Sacks names the moment he thinks will be remembered: the president calling in during Jensen Huang's talk. His argument is that Huang's content and demeanor, more than any argument, did the political work. The president, he says, "did what he did best, which is seize total control of the narrative." That claim is paired with a stronger one — that the weeks before the summit were not an organic panic but a coordinated operation: "there has basically been the SIOP for the last few weeks leading into the summit to convince everyone that humanity was gonna go extinct unless we did exactly what the Democrats want and stop all AI development." Sacks offers no evidence for the psyop framing beyond the sequence of events; he simply asserts that Satya Nadella calmed people down first and Huang did the same afterward.

Friedberg supplies the mechanics, and they matter because they undercut any suggestion of scripting. Jensen was texting the president backstage, told him to call back, and the call arrived while he was on stage. Friedberg calls it "completely spontaneous," and Jason adds the detail that he was the one who put the call on speaker and handed over a spare microphone he keeps on set in case someone's audio fails.

Jensen makes the greatest chips. Everybody tries to reverse engineer them. Impossible. Impossible. But it can't figure out how to press the speakerphone button.

The joke lands, and Jason coins a word for the moment — "Trumptelligence" — but the substantive framing comes in his summary of the previous week: the "nerds were losing their mind" and then "the adults came. Daddy came." Elon Musk, Huang, and the president are cast as the adults; the people warning about existential risk are cast as children who needed supervision. That casting choice is the spine of the first half of the episode, and it is asserted rather than argued.

Chamath Palihapitiya's contribution is the most analytically useful move in the segment. He separates the industry into two cohorts. The first is "up and coming organizations that want to call themselves labs, but are in fact companies that have p and l's and shareholders and have to abide by the same rules that everybody else does." The second is older firms that "have lived in scrutiny for many decades and have learned the hard way the process of making things robust and the cost of making things not robust."

He then names the coalition he thinks formed around that distinction: Satya Nadella, Jensen Huang, Sacks, Mark Zuckerberg, Elon Musk, the president, and Lina Khan. His point is that product liability is the mechanism that already exists to discipline the second cohort, and that the first cohort is asking to be exempted from it by relabeling itself. "We have to stop calling these companies labs. They're companies," he says, and then: "And we have to be judged on the same standards, including product liability."

The concrete evidence offered is revealing because it is entirely about shipping discipline rather than about catastrophe. Musk delayed full self-driving despite claiming he could have shipped it years earlier, on the stated grounds that every Tesla accident is magnified relative to competitors. Zuckerberg, per Chamath, slow-rolled Muse for a couple of months with Alexander to get it right. Neither example involves an existential threshold; both involve the ordinary cost of being a company that can be sued.

Sacks takes the same material in a different direction, toward a contrast between individual and collective responsibility. His formulation is that the operative question is always "should we ship this product?" and never "can we get the whole world to agree?" The failure mode of the collective version, he argues, is that it "defrays accountability away from the actual decision maker."

He aims that at Dario Amodei and Sam Altman appearing at the United Nations to call for global AI governance, and the comparison he reaches for is deliberately unflattering: billionaires flying private jets to Davos to rail against climate change. Jason supplies the one-word version twice — "Hypocritical" — and Sacks adds "Delusional." His conclusion is that a global framework is both a distraction and a delay: "if you guys don't get this decision right, it's not going to matter what the United Nations does," because by the time the UN agrees on anything the technology will have moved well past the agreement.

This is a coherent position, but it is also convenient for the person holding it. Sacks is arguing that the correct locus of decision is the firm, at the same time as he is arguing that the firms are over-regulated in the specific sense of liability exposure. The two claims can coexist, but the segment never stress-tests them together.

Jason then floats what he calls a theory: that the frontier companies are quietly seeking the kind of liability shield the internet industry obtained through Section 230, and that inside the administration there is consideration of a trade in which the companies hand roughly ten percent of their equity to a United States sovereign wealth fund in exchange for relief. His analogy is pointed — it would be like letting an aircraft manufacturer off the hook for planes falling out of the sky because it gave the government a slice of equity.

Sacks resists the framing twice. First he objects that Jason is describing the idea as the administration's, which he says is "not true"; Jason clarifies that the pitch is coming from the companies and that some officials may be receptive. Then, pressed directly on whether the companies are lobbying for protection, Sacks gives the most carefully hedged answer in the episode: "This is not, like, first person experience. They they have not asked me, like, hey. We want product liability waiver, but I have heard from people in Washington that they have been seeking something like that."

What he treats as documented is the response. He says that in the past week virtually every administration official said some version of the same thing — that AI companies must take responsibility for their own products and that product liability will not be waived — and he attributes the line to Speaker Johnson and to Scott Bessent in a recent interview. He extends it to antitrust: no waiver for anyone "trying to form a cartel" or "trying to collude." He also cites a Trump post pointing at the Justice Department, the administrative state, and civil and criminal suits as the guardrails that force a correct risk-reward calculation.