AI Bubble Pops, Zuck Freezes Hiring, Newsom's 2028 Surge, Russia/Ukraine Endgame

The episode opens with banter about Moose, described as a bulldog belonging to one of the hosts' households. Moose is moving to live on a ranch belonging to another member of the group, which one participant jokingly calls an "empty nest" moment. The move is a trial: Moose lands "next week" for a three-month trial and stays if he behaves.

The hosts praise the dog as well trained and "very athletic," not a couch potato, and say he needs room to run. Someone asks why the owner doesn't buy Moose a ranch of his own, and the reply is: "Why buy a ranch when the bestie has won a ranch?" The ribbing turns into comparisons between bulldog traits (stoic, loyal, stubborn, "resist training") and one of the hosts, whom the group calls "Jason" and "Jayco." The joke ends on "founder market fit" recast as "owner pet fit" and "Founder bulldog fit."

An anecdote illustrates the stubbornness. Moose would jump on the couch, get taken down, and jump right back up, again and again, until the family "just give in and let him sit there." The conclusion: "Now it's his couch."

The main show opens by welcoming the panel. The participant addressed as the "chairman dictator" is enjoying the end of summer and argues that a two-to-three-week August vacation should be "mandatory around the world." The host describes a trip to Mexico City as a mix of Spain and Williamsburg, Brooklyn: "super hip, super fun, lots of young people, lots of energy," with a strong food scene and an "amazing" startup scene.

The host then turns to Friedberg (transcribed variously as Freiburg and Freiberg), who spent August working on the group's AI summit and All-In summit. Asked about Ohalo, the reply cites a customer visit that day and "The world's first true potato seed," with production ramping up. A joking exchange follows about who did and didn't get to invest. One participant learns they are already a shareholder. Another complains about missing "a slice" and ends with "good luck with your mutant fucking potatoes."

Two of the hosts recap an interview with Missouri senator Eric Schmidt (as transcribed), a former state attorney general. The source credits him with bringing the lawsuit it calls "Biden v Missouri." In the speaker's telling, the suit exposed a Biden administration "censorship apparatus" run through social media companies, the "government side" of the Twitter files. The speaker says the suit's discovery, combined with the Twitter files Elon released, shows the government "was in bed with social networks" to censor Americans' speech on COVID and other issues. This is the speaker's characterization. The episode presents no further evidence for it.

The rest of this segment promotes the upcoming summit, which the host describes as the fourth one, running September 7 to the ninth with the eighth and ninth as the main days. Named speakers and partners include:

Speakers: Alex Karp, Vlad from Robinhood, Energy Secretary Chris Wright, the Uber CEO and the Circle CEO. Party headliners: Diplo, Gary Richards and Dylan Francis. Solana: returning as a bigger sponsor. The speaker says it is doing "over a 100,000,000 transactions a day" and will run a Sunday open bar, juice bar and coffee. OKX: returning, building on stage, with a matcha bar. Google Cloud: a mini campus and "350 k in credits" for startups who attend. Athletic Brewing: funding scholarships again, with a few slots left. Attendees who can afford a regular ticket are asked not to apply.

Before the main topic, one host mentions appearing in a new South Park episode about "the sycophancy of AI." The episode showed a line of people bringing tributes to the president, inspired by the Apple gift from Tim Cook. A participant praises the administration for taking the mockery in good humor, saying JD retweeted it. That prompts teasing over the casual "my guy JD" for the vice president.

The host frames the week as a "detour" for AI mania. AI stocks fell across the board over three or four days, and the host gives three causes: a viral MIT study, Sam Altman's comments about a bubble, and Zuck freezing AI hiring after a "complete blitzkrieg."

The study, which Fortune surfaced and MIT had published the previous month, reported these figures:

Ninety-five percent of Gen AI pilots fail to reach production. The stated causes are employee resistance, poor-quality output and, the host's "most interesting" one, resource misallocation. 70% of Gen AI budgets go to sales and marketing tools, which the study links to poor ROI. The highest ROI comes from back-office optimization, such as automating tasks that cut departmental spending. The sample was 300 AI implementations, plus interviews with 150 leaders across 52 companies.

The discussion goes no further into the study's methodology than these counts. The figures are presented as the study's findings, not checked independently on the show.

The first answer comes from the participant asked whether the study matches what they see at their software company, eighty ninety. They call the first wave of enterprise AI a top-down reflex. Boards read about AI, asked CEOs "what's your AI strategy?", and the question moved down to the CTO. Companies with large budgets then "try a bunch of different things." The speaker says a sorting phase is now underway, built on recognizing "that there's a big difference between probabilistic software and deterministic software."

That distinction is the speaker's explanation for both of the study's main findings. Sales and marketing fail because they are hard to codify "into a set of heuristics that never change." Back-office work suits AI because many employees there are hired to handle edge cases within a defined process. The speaker says that done correctly, back-office AI can reach "extremely high rates of accuracy."

The speaker's bigger open question is how durable the revenue is. They describe companies reaching large ARR in months (the transcript renders the figure as "$50,100,000,000 dollars of ARR") and then raising huge rounds. What hasn't been seen yet is logo churn or dollar churn, which could come from three directions: cheaper new entrants, foundation models "move up the stack and just absorb capability", or products that fail and get abandoned.