Leo Laporte opens the show with a preview of the major stories — Anthropic's Fable ban, OpenAI's financials, and the SpaceX IPO — before diving straight into an extended pre-show warmup about the 2026 FIFA World Cup hosted in the US. Doc Rock, a longtime football fan who played until his knees gave out, waxes lyrical about Dutch supremacy and the electric atmosphere at live games, especially in Japan. Iain Thomson shares his delight at Scottish fans drinking a transatlantic flight dry of beer. Owen Thomas notes San Francisco bars going World Cup-crazy. The hosts debate American attitudes toward soccer versus American football, with Doc Rock delivering a spirited defense of the beautiful game.

The show's main story kicks off with Leo detailing how Anthropic's Fable model — essentially its powerful Mythos model with safety classifiers — was banned by the US Commerce Department after an Amazon research team demonstrated it could not only find software vulnerabilities but write test exploits to verify patches. Security researcher Katie Mazuris reviewed the Amazon paper, but her close association with Chris Krebs — the CISA director Trump fired for certifying the 2020 election — may have been the real political trigger, according to TechDirt's Mike Masnick. Iain Thomson notes that security expert Alex Stamos launched freefable.org with prominent signatories arguing every major AI can do what Fable does. The ban, Leo argues, is legally questionable, practically unenforceable, and mirrors the failed 1990s encryption export ban.

Owen Thomas delivers the sharpest political insight of the episode: OpenAI's chief global affairs officer Chris Lehane is the same mastermind who navigated Airbnb through city-by-city regulatory wars, and Anthropic needs to 'staff the F up in Washington.' Leo adds that OpenAI's Greg Brockman donated millions to Trump's PAC, OpenAI pledged 10% of its IPO to the US government, and Sam Altman stood next to Trump at the inaugural for Stargate. Anthropic did none of this and got banned. Doc Rock invokes the prohibition analogy: banning AI, like banning alcohol or drugs, will drive it underground and ultimately fail. The conversation turns to whether Trump may simply be reading polls showing that 61% of Americans don't want AI in products at all.

Leo delivers a warm, personal Shopify ad read centered on how his son Henry built a successful food business using Shopify. He highlights the platform's ease of website creation, marketing tools, and the iconic purple Shop Pay button, describing it as the best-converting checkout in the market. Listeners are directed to shopify.com/twit for a $1-per-month trial.

The conversation pivots to SpaceX's recent IPO and what Elon Musk's empire of companies actually means. Owen Thomas explains that SpaceX's $60 billion acquisition of AI coding startup Cursor was partly paid in SpaceX stock — and that SpaceX shareholders actually paid less as a percentage as the stock rose. The New York Times reports that Musk's next move may be merging SpaceX and Tesla, driven by Tesla's chip manufacturing ambitions in Austin. Iain Thomson has read the SpaceX S-1 and delivers the line of the segment: it 'makes WeWork look optimistic.' The discussion covers why Musk moved his companies to Texas — Delaware courts had already struck down his trillion-dollar pay package — and the impracticality of the Mars colony central to SpaceX's pitch, citing Kelly and Zach's book 'A City on Mars.'

Ed Zitron's independently verified documents, published by the Financial Times, show OpenAI lost $38.5 billion last year — but Owen Thomas provides essential context: much of the loss reflects non-cash restructuring from OpenAI's nonprofit-to-for-profit conversion, along with stock-based compensation. The company now has approximately one billion weekly active users and filed confidentially to go public. Meanwhile, Google suffers a massive talent blow: Noam Shazeer, co-author of the landmark 'Attention Is All You Need' transformer paper and architect of Google's Gemini models, is leaving for OpenAI — just two years after Google paid $2.7 billion to hire him via the Character AI acqui-hire. The panel speculates wildly about what OpenAI could possibly have offered him. Doc Rock wishes Anthropic would better explain its constitutional AI approach to a skeptical public.

Leo brings the data that reframes the entire Fable ban story: according to RAMP, Anthropic's share of paid AI business subscriptions rose 2.5 percentage points in May to 41%, beating OpenAI's flat 39.5% — and the timing suggests the ban itself was the catalyst. Owen observes dryly that 'every time they get banned, they get more customers.' Anthropic had long been the insider's choice for software engineers but lacked OpenAI's name recognition. The ban changed that instantly, turning Fable into forbidden fruit.

Leo pivots to a sponsored segment for Simply CX, Microsoft's customer experience podcast hosted by Nicole McKinley, highlighting an airline contact center case study that reduced average handle time by 30% using AI. He then delivers a personal Ethos life insurance ad tied to Father's Day — reflecting on becoming a parent and the immediate instinct to protect his family — directing listeners to ethos.com/twit for same-day coverage with no medical exam.

With the Anthropic story as backdrop, Leo steers into the deeper question of how AI should be used. He shares his own experience removing his Oura Ring after realizing the device was mediating his relationship with his own body. The same applies to AI and writing: 'the reason you learn to write is it's a process of learning to think.' Iain Thomson cites a teacher shocked by students blindly submitting Claude's output without checking it, and a UK newspaper journalist who published 110 articles in a single day using AI. Doc Rock offers the counterpoint: he does morning pages by hand every day, then walks and dictates to Claude, which synthesizes his ADHD brain dumps into prioritized notes. The panel agrees the answer is responsible adoption, not prohibition — and that tech illiteracy itself is a national security risk.

Leo reports that Tim Cook told the Wall Street Journal that Apple will have to raise prices due to memory chip shortages — kicking off what he expects will be a months-long campaign before September's iPhone launch. The panel traces the chain: data centers are consuming memory at scale, tariffs closed the Strait of Hormuz, GPU and RAM prices have surged, gaming hardware became unaffordable, and young people are furious. Nothing's Carl Pei cancelled the CMF budget phone because RAM now costs double and represents more than half of total phone build cost. This economic pressure, Leo argues, is one real driver behind public anger at AI companies.

In a lighter interlude, Iain Thomson explains he's going through US citizenship biometrics — fingerprints, eye scans, and a facial scan — and studying for the civics test. Leo reminds him that knowing the US Constitution in detail makes him better prepared than most Americans. The conversation meanders back to soccer: Doc Rock describes watching Johan Cruyff and Pelé play at RFK Stadium as a child, which cemented his lifelong devotion to the Dutch national team. The orange mic muff is now explained.

Leo went to see Bob Dylan — his first time in 50 years — at the Greek Theatre, where signs warned that phone use meant ejection. Despite this, audience members were sneaking their phones constantly, proving how addictive the devices have become. Iain Thomson was publicly shamed by Stephen Merritt of the Magnetic Fields for tweeting during a concert. The panel debates screen addiction, its particular harm to teenagers, and whether banning phones at concerts is reasonable. The conversation shifts into online privacy: Leo discovers that even his privacy-focused Zen Browser has a unique fingerprint, Doc Rock details how many trackers appear on a simple news article, and Leo reveals that TWiT was hit by a sophisticated man-in-the-middle phishing attack — including stolen two-factor authentication credentials — that went undetected for 121 days.

Leo and Iain dig into the UK government's announcement that children under 16 will be banned from YouTube, TikTok, Facebook, and Snapchat by the end of the year. The plan's enforcement mechanism is murky at best — it appears to rely on facial recognition age checks or credit card verification, neither of which will stop determined teenagers. Iain, who has friends in Australia, confirms the equivalent ban there has not worked. He also raises the privacy nightmare of requiring government ID for all social media users. The panel agrees that not allowing phones in schools would be a far more practical first step, noting that Apple has already built parental control defaults into iOS based on American Pediatric Association research.

Leo explains Meta's sophisticated lobbying play: support the Kids Online Safety Act in a form that transfers age-verification liability to Apple and Google app stores, while shielding Meta from over 2,000 active lawsuits. The company faces cases from families, school districts, and state attorneys general following a landmark Los Angeles jury ruling that Instagram's algorithm constitutes a defective addictive product — analogous to cigarettes — rather than merely a platform with user-generated content. That legal theory sidesteps Section 230. Iain Thomson summarizes it bluntly: 'You've legalized bribery and called it campaign contributions.' Leo argues for keeping Section 230 while still holding Meta liable for algorithmic design choices, calling them genuinely contradictory positions that the law needs to resolve.

Leo gives Apple credit for its WWDC child safety announcements: rather than leaving parents to configure complex settings, Apple now ships defaults designed in consultation with the American Pediatric Association, making it easy for parents to simply 'turn it on.' More importantly, Apple created an API mechanism that allows any app to ask the iPhone for a user's age group and adjust behavior accordingly — a cleaner solution than having thousands of apps build separate age-verification systems. This is exactly what Apple's regulatory critics have argued for: leverage the app store as the enforcement layer. John Gruber's caveat is also raised: some of this is getting ahead of government regulation rather than genuine child advocacy.

Leo explains Box's pitch for the AI era: public AI models know public information, but they don't know your company's product roadmaps, sales materials, or financial models. Box serves as a secure content layer that enterprise AI agents can query, with tools like Box Agent, Box Extract, and Box Hubs. He highlights a stat from Box's State of AI in the Enterprise report: 96% of organizations say agents need company-specific data, but only 36% have connected them to trusted content. Listeners are directed to box.com/AI.

Leo details the just-unsealed story of Google's fight against a DOJ keyword warrant issued in the investigation into pipe bombs planted near DNC and RNC headquarters after the January 6th Capitol riot. Google initially complied with a narrower warrant identifying 250+ users in the vicinity, but fought a second, broader warrant covering anyone who searched for the committee names combined with words like 'bomb' or 'explosive' — potentially affecting 300+ entirely innocent users. US District Chief Judge James Boasberg, citing the magistrate court's earlier ruling, ordered compliance. The panel discusses how AI could make such mass data requests more tractable for investigators, potentially incentivizing broader warrants — Iain's dark quip: 'So now our freedom is based on ChatGPT.'

Leo reports that Waymo issued its sixth recall — this time removing nearly 4,000 robotaxis from highways — after 13 documented instances of vehicles driving through lights, barricades, and barriers into closed construction zones. Six incidents occurred in Phoenix in April; seven in San Francisco in May. This is far from Waymo's first recall: previous ones addressed flooded road behavior, illegal maneuvers near school buses, and low-speed collisions with chains and gates. Owen notes that Waymo is now testing in London. Despite the recalls, Waymo maintains it has driven more than 170 million autonomous miles with a 13x reduction in serious injury crashes versus human drivers — a claim the panel largely credits, noting that human drivers are terrible.

Leo delivers the Zscaler sponsor segment, focused on AI-era security risks: generative AI is a boon for threat actors creating phishing lures and malicious code at speed, and employees are inadvertently leaking sensitive data to AI applications. A Zscaler stat: 1.3 million Social Security numbers have been leaked to AI apps. BioIVT's acting CISO Chad Pallett testifies that Zscaler helped them reduce cyber insurance premiums by 50% while doubling coverage. Listeners are directed to zscaler.com/security.

Leo reads aloud from the UK government's own fact sheet for the social media ban and finds it bewildering: children can still use WhatsApp and Facebook Messenger, but not YouTube. Adults who have had accounts for more than 16 years or have a credit card linked may be exempt from age checks. Iain Thomson — who has lived in both countries — calls the legislation 'ridiculous' and points out it was modeled on Australia's ban, which has not worked in practice. Doc Rock draws the historical parallel to online predator scares on AOL and CompuServe in the 1990s, arguing the problem wasn't the platform but human behavior, and no platform restriction has ever solved it.

Leo introduces Snap's Spectacles X — the self-contained AR glasses Snap is spinning into a subsidiary called Specs Inc. The panel immediately mocks their appearance (Iain coins 'birth control glasses'), but Doc Rock argues that chunky frames are actually in fashion right now thanks to K-pop. The glasses feature a heads-up display roughly 2-3x larger than Meta Ray-Ban's, 4 hours of battery life with four charges from the case, and full self-containment without a smartphone dependency. Leo notes that Snap's stock is down over 95% from its 2021 high, making this a desperate, all-chips-in bet. Everyone agrees they're waiting for Apple, whose iPhone processor offloading could solve the weight and battery problems in a way Snap cannot.

Leo breaks the Fox-Roku deal, describing it as a play for the post-broadcast future: Fox owns Tubi, a major FAST service, and by combining it with Roku's 80+ million connected TV platform, it becomes one of two dominant free streaming players alongside Paramount's Pluto. Doc Rock delivers the most visceral critique: trying to watch the World Cup on the Fox Sports app requires scrolling past seven rows of Fox-branded content before reaching the games, and every ad break is filled with unsold inventory featuring Jesus commercials and America-first programming. Owen contextualizes the broader trend of media consolidation around FAST services, which are proving more popular than pure subscription tiers among mainstream viewers.

Triggered by the Snap Spectacles discussion, Doc Rock reveals his real-world workflow with Meta Ray-Ban Gen 2 glasses: he uses them primarily as ambient audio for podcasts during travel and as a microphone for walk-and-talk voice memos fed into WhisperFlow and AudioPen, which Claude then synthesizes into organized notes. His secondary use case is reading Japanese text he can't parse on restaurant menus in Kyoto and Osaka — glasses-based real-time translation. He deliberately chose the model without the heads-up display to save weight. The panel agrees that tying smart glasses to an iPhone processor would be the game-changer Apple hasn't yet delivered.

Leo reveals Midjourney's stunning pivot: the company known for AI image generation is building full-body ultrasound scanners using 40 Butterfly Network chips per unit, analyzed by Midjourney AI, and opening a spa in San Francisco's Union Square before end of 2027. The goal is to compete with on-demand MRI services like Prenuvo at a lower price point. Owen Thomas notes that Union Square is increasingly an AI experimentation hub — Worldcoin's orb is nearby — and the specific location on Grand Avenue is some of the poshest real estate in the city. Leo frames it as Snap-style 'pushing all your chips to the middle': a high-risk, high-reward pivot from a company betting everything on a new market.

Leo reports that Nothing founder Carl Pei cancelled the CMF phone because RAM prices doubled, making memory the single most expensive component in the phone — over half the total cost. This isn't isolated: a 20TB hard drive that Doc Rock bought in December for $289 cost $700 for the exact same model a few months later. Western Digital reportedly sold out its entire annual hard drive production months in advance. The panel traces the chain: AI data centers are consuming memory at scale, tariffs and the Strait of Hormuz closure disrupted supply chains, and the result is that consumer electronics — from budget phones to gaming GPUs — are becoming unaffordable.

Owen Thomas, whose publication ran the cover story, explains that the AI IPO wave is unlike previous tech booms: OpenAI and Anthropic grew so fast that their early employee cohorts are far smaller than at Facebook or Salesforce, meaning the wealth created by going public will be highly concentrated. SpaceX's IPO minted roughly 400 millionaires, not the thousands that Google's IPO created. Meanwhile, both companies are near or at one million square feet of leased office space in San Francisco — Anthropic has cornered three consecutive blocks of Howard Street (dubbed 'Anthropic Row'), and OpenAI has dominated Mission Bay. The panel compares this to Amazon and Microsoft transforming Seattle, and to Salesforce's earlier role in San Francisco's commercial real estate — though Leo notes Twitter's transformative moment didn't last.

Leo, Owen, and Doc Rock spiral into the biggest structural questions of the AI era: Will AI eliminate jobs or transform them? Owen argues AI will raise the bar for education and critical thinking, not lower it. Doc Rock argues that tech illiteracy is more dangerous than any AI model, enabling misinformation and poor policy decisions. He recalls how NASA parents donated the computers that got him hooked on technology in the 1970s, and draws a line from Reagan-era deregulation in 1984 to today's income inequality explosion. Iain Thomson cites George Carlin's sketch about rulers wanting people 'educated just enough to push the buttons' — and observes it's aging remarkably well. Leo closes by pointing out that an educated electorate is bad for politicians who depend on fear campaigns.

Leo pays tribute to two figures who died in separate plane crashes: Joshua Baer, founder of Capital Factory, who is credited with turning Austin into a genuine tech hub and was widely beloved in the startup community; and Claude Guillemot, co-founder of Ubisoft, who with his brother transformed a mail-order software business into one of the world's largest video game companies. Leo then delivers the Club TWiT pitch, highlighting upcoming programming including Jeff Atwood's 'Off by One' show and the AI User Group. The panel offers World Cup predictions — Doc Rock backs the Netherlands, Iain Thomson is hoping for Scotland but expects France or the Netherlands, Owen just wants watch parties to fill San Francisco bars. Leo signs off with Father's Day wishes and a reminder that TWiT records every Sunday from 2–5 PM Pacific.