Thomas Laffont: IPOs and SPACs are Back, Mag 7 Showdown, Zuck on Tilt, Apple's Fumble, GENIUS Act passes Senate

The episode opens with the usual banter. The hosts congratulate David Friedberg on his fourth child and swap jokes about a gift basket of picanha steaks from Long Hill Wagyu. They complain that after an earlier on-air mention, listeners bought out the ranch's whole stock. The group introduces Chamath Palihapitiya as "chairman dictator" with "two votes" and welcomes guest Thomas LaFonte. David Sacks is absent because he is at the White House.

LaFonte hosts East Meets West, a conference that one of the hosts attended that week. Asked for highlights, a participant says AI news dominated most panels "pretty much up and down the stack": SaaS companies trying to transform into AI businesses, Meta's deal with Scale, and a report in The Information about Nat Friedman. Rick Caruso also appeared at the conference, which was held at his property. The hosts call him "the mayor who would have saved Los Angeles from the fires."

The conversation turns to the state of Los Angeles. One participant cites a chart a friend had sent that day: per store location, restaurant recovery after COVID in LA is 50% behind the national average. Two explanations are offered. The first is concentration. San Francisco's economy is "levered to to AI and on the upswing," while LA's is tied to entertainment, which the speaker describes as in secular decline. Someone at the conference said filming in LA is down 50% from its peak, as production moves to Georgia, New York, the UK, Atlanta and other hubs that offer "much better deals than Los Angeles is."

One participant gives a first-hand example from MrBeast's Beast Games. That production got large tax credits for shooting in Las Vegas and Toronto. Its second season for Amazon rests on "an enormous deal with the Kingdom Of Saudi Arabia," where the sets are being built and will stay. The speaker is blunt about California:

We would not film in Los Angeles unless we absolutely had to. We will stay as far away from California as possible.

The second explanation is regulation and cost. Participants mention a figure of shooting in LA being "30% more expensive," which one calls "kind of the official number," though the phrasing is hedged. They also point to the time and paperwork needed to get a production running. Research from the James Beard Foundation is cited as finding that independent restaurant owners cannot get staff. One speaker concludes that people find it hard to live in LA and to get through its regulations, and calls the idea that California has a lock on anything beyond "incredible weather and beautiful people" "farcical."

Another conference theme was the US debt-to-GDP ratio, approached from the GDP side. The question raised was whether AI could raise productivity and grow GDP faster than expected. That, the speaker suggests, might be "one of the reasons why" interest rates may not be as high as recent trends would imply. The time frame given is the next "five to ten years." One participant argues the benefit should land first in the US rather than in global GDP. Another asks whether AI-driven growth depends on bringing outsourced manufacturing back onshore. The reply is that onshoring is "part of it," and the other part is productivity gains among knowledge workers.

The concrete evidence offered is the company OpenEvidence (transcribed as "open evidence"), a diagnosis engine. A participant says "a third of US physicians are on the platform using it, 10 times a day," with notable traction in oncology. The speaker then extends the pattern to law and coding and imagines "an explosion of productivity gains" across the physical and digital economy. This part is framed as a question rather than a measured result.

Another participant spells out the economics using preventive checkups. Today these are expensive, appointments are hard to get, and insurance may not cover them. If AI lets a doctor's throughput rise by 10x, the price per checkup can fall while more people get checked each day. The speaker argues this grows the diagnostic-care slice of GDP. The general principle offered is "Anything where AI provides leverage to a service provider where their throughput now goes up."

A lighter example follows. An LA dentist went viral with an ad made with Veo 3 (transcribed as "VO3") about a skydiving gorilla who needs dental work. According to the speaker, the ad "probably" cost a couple hundred dollars to make, and the practice is now full and flooded with requests for implants. After this segment, the host restarts the show and introduces the docket.

The first docket item is the framing that "Zuck is tilted." According to reports, Mark Zuckerberg is frustrated that Meta is falling behind in AI. The host relays Sam Altman's claim that Meta offered top OpenAI employees a $100,000,000 signing bonus and also offered $100,000,000 a year in annual compensation. The host adds, "Who knows if this is true or not?" The host compares the tens of billions apparently set aside for this push to Meta's VR efforts, which "didn't work out so well."

The show plays a clip of Altman on his brother Jack's podcast, Uncapped. Altman describes "$100,000,000 signing bonuses, more than that comp per year" and says he is glad that "at least so far, none of our best people have decided to take them up on that." He explains why:

OpenAI has got a really good shot, a much better shot at actually delivering on superintelligence, and also may eventually be the more valuable company.

The host then describes Meta's investment of over $14 billion in Scale AI for a 49% stake, putting "invested" in quotes. The host calls it "a shadow aqua hire" (acqui-hire) designed to avoid antitrust scrutiny. The cited precedents are Microsoft with Inflection AI, Google with Character AI and Amazon with Adept (transcribed as "the depth AI"). The host doubts the structure is still needed now that Lina Khan is no longer in her post. Scale CEO Alexandr Wang and others will join a new Meta superintelligence team. Scale says it will remain independent under a new CEO, which the host is "not sure" will happen. Scale supplies data labeling, using experts to help train language models. The host says two of its biggest customers, OpenAI and Google, both canceled their contracts, and describes Meta as taking "that chess piece off the board."

Meta is also reportedly in talks to hire former GitHub CEO Nat Friedman and Daniel Gross, who run an AI incubator and investment fund. Meta has $70 billion in cash. The host asks how big the prize of having the best large language model is, and what to make of these aggressive pay packages and 49% purchases.

One response calls the spending "highly rational" and does the arithmetic. Meta's market cap is roughly $1.7 trillion. If the CEO believes maybe 50% of it, about $850 billion, is at risk from AI, then spending four or 5% of that is justified if it "increases the odds even slightly" of winning. The same speaker suggests speed may matter as much as regulatory cover. Structured this way, Wang could start at Meta "literally, the next day."

The speaker compares the Scale deal to Facebook's purchase of Onavo. Onavo ran a panel of phones that let investors see which apps people actually used. It was "the only service that gave you true engagement data." After Facebook bought it, the company used it internally and cut off everyone else, so investors lost a key way to tell who was winning in the mobile app era. The host summarizes the parallel: Meta buys a service many rely on, shuts it to others, and keeps the tool and the data for itself. The guest agrees and adds that, given Meta's market cap and the size of the opportunity, the move makes sense.

The comparison raises an older Facebook debate: building mobile apps as HTML5 wrappers or as native apps. A participant who sat on Facebook's executive team says they were "on the side of apps" while another executive backed HTML5. The participant says their political capital was tied to native apps and to "our own phone, an entire verticalized integrated stack," and describes their own style then as "scorched earth, get it done."

The participant's case was that by 2010 native apps were obviously right. HTML had one legitimate use: getting around carriers and ecosystems that tried to charge a toll. In 2010 the speaker's team built a URL-only replica called Facebook Zero and launched it at Mobile World Congress. In markets such as India, carriers tried to tax Facebook, but users who went through the browser avoided the charge. The speaker presents this as the right tactic for developing markets and a side effort only. The real solution was "an extremely integrated app from the software all the way to the hardware."

In the participant's account, the decision instead "became politicized" and turned into a big bet on HTML5. They left after concluding the phone project would not happen. About a year later, they say, Zuckerberg called the bet a mistake, "ripped all the HTML five stuff apart, went native." Another participant adds that open-standards idealism among some developers also drove the choice. The first speaker rejects that view and uses slurs and profanity for its advocates. The speaker sums it up:

people politicizing what should have been an obvious technical decision.