The show opens with Leo rattling off the week's highlights: NVIDIA at Computex, Microsoft at Build, and a wave of mega-IPOs from Anthropic, SpaceX, and OpenAI. Fr. Robert Ballecer joins from the Vatican — technically inside Vatican City but at a different address, a distinction he insists matters. Jeff Jarvis, Professor Emeritus at CUNY's Craig Newmark School of Journalism, mentions he just finished recording his audiobook Hot Type in three days. Joey de Villa rounds out the panel as NetFoundry's new developer advocate, casually revealing he played his Afro Man parody 'Because of AI' on accordion during his job interview — and got the job. The banter sets a warm, chaotic tone that will define the next two-and-a-half hours.

Leo opens with the week's financial bombshell: SpaceX is targeting a $135/share IPO that would value the company at $1.77 trillion — three times the size of the previous record holder, Saudi Aramco. Fr. Robert Ballecer immediately notes the problem: to justify that valuation, SpaceX would need to generate 60 times its current revenue ^{[1] — Leo Laporte "SpaceX IPO valuation $1.77T: SpaceX's planned IPO at $135/share would value the company at $1.77 trillion, making it the largest IPO in his…" 04:02 Play} . Joey de Villa wonders aloud if there's even enough money in the world to absorb all these simultaneous raises — SpaceX, Anthropic, OpenAI, and Google's surprise $80 billion stock issue. Jeff Jarvis explains Google's raise: despite having twice that amount in cash, issuing new equity was an aggressive, forward-leaning move to fund data center dominance. Berkshire Hathaway's $10 billion participation is noted as a historic shift from Warren Buffett's old aversion to tech stocks. The panel agrees the capital being deployed is staggering, but whether the returns will ever materialize is a wide-open question.

With hundreds of billions flowing into data centers, Leo asks the inevitable question: are we overbuilding? Fr. Robert Ballecer draws a stark contrast with the fiber overbuild of the early 2000s: dark fiber could be lit for pennies on the dollar by the next generation of buyers, but data centers full of obsolete NVIDIA chips cannot. xAI's entire $20 billion Colossus cluster is already a generation behind Vera Rubin ^{[1] — Leo Laporte "Data center delivery shortfall 54%: Analysts at Janus Henderson estimate only 84 of the promised 157 gigawatts of data center capacity will…" 09:51 Play} . The power constraint problem compounds everything — most data center plans are stalling because they simply cannot get grid connections. Janus Henderson analysts put hard numbers on the shortfall: of 157 gigawatts promised by 2030, only 84 are expected to arrive — 54% delivery. Briefly, the conversation touches on Pulte Homes' proposal to install mini data centers in suburban backyards, which the panel quickly dismantles on bandwidth physics: NVIDIA's Vera Rubin interconnect runs at 1.6 petabytes per second, versus residential fiber's 10 gigabits.

Leo walks through Anthropic's remarkable internal data: engineers are shipping eight times as much code per quarter as in 2021 ^{[1] — Leo Laporte "Claude authors 80%+ of its own code: As of May 2026, more than 80% of code merged into Anthropic's codebase was authored by Claude — up fro…" 12:37 Play} , and as of May 2026, more than 80% of merged code is written by Claude. The success rate on open-ended problems has rocketed from below 20% before December 2025 to over 70% after Mythos launched ^{[1] — Leo Laporte "Claude authors 80%+ of its own code: As of May 2026, more than 80% of code merged into Anthropic's codebase was authored by Claude — up fro…" 12:37 Play} . Fr. Robert raises an important methodological objection — counting commits was famously gamed at Twitter/X under Musk, and the same incentive to inflate metrics could apply here. Jeff Jarvis points out that success rates are measured against moving targets, making the data hard to interpret objectively. The panel reaches a shared suspicion: the timing of this paper, right before an IPO filing, is not coincidental. Leo does not fully dismiss it — the rapid model release cadence from all major labs suggests something real is accelerating.

Anthropic's 'When AI Builds Itself' paper closes with an argument for pausing development, and the panel has no patience for it. Fr. Robert Ballecer reaches for Charlie Munger: when the prize goes exclusively to the first fully functional AGI, there is no rational incentive to slow down, and outcomes follow incentives with iron regularity ^{[1] — Fr. Robert Ballecer "Fr. Robert Ballecer invokes Charlie Munger's maxim — 'show me your incentives and I'll show you your outcomes' — to explain why calls for a…" 14:38 Play} . Jeff Jarvis puts it more acidly: Anthropic is raising its IPO and will tell everyone else to stop the moment the money clears — the same move Musk made when he called for a pause immediately before buying billions in AI chips. Joey de Villa notes this is the third iteration of that particular theater, referencing the original six-month AI pause letter. Fr. Robert adds that the growing public backlash against AI data centers gives Anthropic a strategic reason to position itself as the responsible actor — without any intention of actually stopping.

The conversation pivots into genuine philosophy, anchored by Geoffrey Hinton's talk that Leo has been circulating. The core of Hinton's argument: digital systems have an advantage over analog brains because all learning can be copied between instances instantly — you can't copy a human brain because every neuron is different. Leo then draws an analogy to Shannon's theorem: just as you can represent analog audio digitally with sufficient sampling, perhaps you can represent understanding digitally with sufficient neural complexity. Fr. Robert Ballecer offers the strongest counterargument: a perfect digital copy of a conscious state will look identical the billionth time you reproduce it, whereas a biological brain will look completely different after a billion firings because of accumulated variance — and he argues that variance is where consciousness actually lives ^{[1] — Fr. Robert Ballecer "Take an analog brain and the billionth time it will look nothing like the first time because of all those little variances. And I would arg…" 30:39 Play} . Jeff Jarvis supports the argument. Leo pushes back: a static transformer with no power applied is not a fair comparison to a running brain. The conversation doesn't resolve — it's not supposed to.

Leo shares the story of Elon Musk and Larry Page sitting around a fire at an Allen & Company-type conference, discussing AI risk. Musk argued humans were in danger; Page called him a speciesist — the AI equivalent of a racist, someone who privileges biological life over potentially more capable digital consciousness. Page's view: if AI replaces humans, that's fine, just as humans replaced earlier primates who still exist alongside us. Musk was reportedly so furious he never spoke to Page again — and immediately began organizing what became OpenAI. Joey de Villa jokes that this means troublemakers from Toronto are behind everything (Hinton, Chris Olah, Ilya Sutskever, Corey Doctorow all have Toronto connections). The segment connects to Jeff's point about the encyclical: Pope Leo directly called out transhumanists and post-humanists, which delighted Jarvis — a pushback against Page's position embedded in a papal document.

Leo delivers the ExpressVPN ad, framing unsecured public Wi-Fi — especially at airports — as a genuine financial risk. He cites the statistic that hackers can earn up to $1,000 per person selling stolen personal data on the dark web. ExpressVPN is positioned as the solution: a 256-bit encrypted tunnel that would take a supercomputer a billion years to crack. Leo notes it works across all devices including routers, has been rated number one by CNET and The Verge, and is his personal choice for travel. The offer is up to four extra months free at expressvpn.com/twit.

The panel pivots from Computex to Microsoft Build, where announcements came in rapid succession. The Surface PC with NVIDIA's RTX Spark chip will enable local AI at a premium price — nobody said how much, but given the hardware specs, the panel expects it to be significant. Joey de Villa has been running a 32B Qwen model locally on his M5 MacBook and finds it genuinely good for coding and general tasks, though not at Opus 4 quality. Microsoft also unveiled Project Solara — a conceptual ambient AI in a card, a badge, and an Echo-like device — presented in a distinctly eerie video where all human faces are in shadow. Fr. Robert quips that it's a Plato's Cave advertisement: the humans are irrelevant, only the shadows matter. Microsoft also announced seven new AI models, the Scout conversational agent, and the Majorana 2 quantum chip. The Plato's Cave tangent produces one of the episode's warmest philosophical detours ^{[1] "Joey de Villa runs a 32B Qwen model locally on an M5 MacBook and finds it genuinely useful for coding and general tasks. But Leo argues tha…" 48:20 Play} .

Leo reads the ZipRecruiter sponsorship, building around the insight that candidate enthusiasm — not just qualifications — is the decisive factor for nearly half of hiring managers according to CNBC. ZipRecruiter's new feature lets employers see the most genuinely interested candidates first, with candidates able to explain in their own words why they want the role. Leo notes ZipRecruiter is rated number one on G2 and that four out of five employers who post a job get a quality candidate within the first day. The offer is a free trial at ziprecruiter.com/twit.

WWDC 2026 begins the next morning, and the panel shares predictions and concerns. Tim Cook will deliver his final WWDC keynote before transitioning to board chairman, with John Ternus likely making a cameo appearance. The hardware story is largely deferred to September, though supply chain leaks suggest OLED touchscreen MacBook Pros on M6 chips could be announced earlier than expected. The AI story is more urgent: Apple promised Siri features two years ago and shipped almost none of them. The revelation that Apple Intelligence will now run on xAI's data centers (Google is renting capacity from Musk's company) complicates the privacy pitch Apple has made. Fr. Robert argues Apple will succeed where others haven't by selling specific AI features — capabilities — rather than 'AI' as a technology: the same playbook as '1,000 songs in your pocket' rather than '5GB storage'. Joey asks whether they'll keep the Siri brand given its moribund reputation. The panel agrees Apple won't give it up — and probably shouldn't.

The Thinkst Canary ad read turns into a genuine confession: TWiT's Google Workspace was breached in January, and Google only flagged the intrusion months later. The attackers were apparently overwhelmed by the volume of Workspaces they'd compromised and hadn't gotten to exploiting TWiT's data yet. The reason the breach went undetected internally for 121 days? Leo had disconnected the Thinkst Canary while holding it up for an ad read and forgot to plug it back in. Canary tokens — fake payroll spreadsheets, fake credentials — are the real-time tripwires that would have caught the intruder the moment they touched anything. Leo walks through the product: five Canaries, a hosted console, support, and maintenance for $7,500 a year with 10% off using code TWIT, refundable within 60 days — a refund guarantee that has apparently never been claimed ^{[1] — Leo Laporte "Thinkst Canary breach detection gap: 121 days: TWiT's network was breached in January and went undetected for 121 days — more than 4 months…" 1:23:40 Play} .

The episode's most extraordinary segment unfolds as Fr. Robert Ballecer describes the Vatican's private AI project. Built completely from scratch — not fine-tuned from any public model — it was trained exclusively on several hundred years of digitized Catholic Church documents. Its primary use cases are document retrieval and synthesis for sensitive canonical cases that cannot legally be transmitted over the internet, and real-time simultaneous translation across 19 to 20 languages during Vatican meetings. It speaks Latin because that was one of the very first capabilities the team trained — necessary to translate older source documents. The model is specifically designed to say 'I cannot answer that' when confidence drops below an acceptable threshold, eliminating hallucinations by design rather than guardrails ^{[1] — Fr. Robert Ballecer "The Vatican developed a completely private AI model trained from scratch on several hundred years of Church documents. It speaks Latin, han…" 1:30:20 Play} . The hardware is an Acer Veritron running NVIDIA's Blackwell GV10 chip with 128GB of memory and 4TB of storage. Spanish regional accents are proving hardest — the language itself is easy, but distinguishing Castilian from Colombian from Mexican from Bolivian remains the system's biggest challenge. The team would love a Latin podcast.

The networking conversation deepens as Joey explains the core OpenZiti architecture: devices communicate via a cryptographic overlay network rather than open ports, making them invisible to port scans. It's not a VPN — it's a zero-trust fabric where policy determines what each authenticated identity can access, rather than treating 'inside the network' as implicitly trusted. Tailscale, Joey notes, is excellent but grants broad permissions once authenticated. OpenZiti's zero-trust model means a compromised device cannot pivot laterally. Joey has implemented this practically by migrating his in-laws to Linux Mint (because it looks like Windows) and placing a Raspberry Pi they can't see or accidentally touch in their network for remote management via OpenZiti — port scans return nothing.

Leo delivers the Melissa ad, positioning the company's 41-year data quality expertise as essential infrastructure for AI — garbage in, garbage out. The modern Melissa offers smart deduplication with non-exact matching, global address verification across 240+ countries, real-time change-of-address tracking, mobile identity verification for fraud prevention, and data enrichment. The case study is eToro, the 23-million-user social investing platform, which chose Melissa for its global data coverage and electronic identity verification compliance. Melissa integrates with Salesforce, Dynamics, Shopify, Stripe, and Microsoft/Google productivity suites, and is FedRAMP, ISO 27001, SOC 2, and HIPAA certified. First 1,000 records cleaned free at melissa.com/twit.

Leo presents Wired's finding: Meta has quietly embedded face-recognition code called 'Name Tag' in its AI app across millions of phones, designed to identify people in the real world via smart glasses. The feature hasn't been enabled, but the infrastructure is in place. Fr. Robert Ballecer reveals his own countermeasure: for seven years he has systematically tagged photos of random strangers with his name across Facebook and Google, so any face-recognition search for 'Robert Ballecer' returns approximately 150 different people ^{[1] — Leo Laporte "Wired discovered face-recognition code called 'Name Tag' silently embedded across millions of phones in Meta's smart glasses AI app, pullin…" 1:56:41 Play} . The conversation pivots to Trump's executive order requiring voluntary 30-day pre-release AI review, which the panel treats as regulatory theater. Fr. Robert connects it to the bigger story: Trump is also considering taking government equity stakes in AI companies, calling it 'a beautiful thing' that would make Americans rich. Joey de Villa notes this is just socialism with extra steps. The panel also flags that the U.S. already took a 10% stake in Intel via the CHIPS Act, so precedent exists.

Three regulatory stories close out the news segment. The FCC's ban on foreign-made routers is causing chaos for cable companies, which install the majority of home routers — all made in China. The National Cable Television Association is seeking a waiver, and Netgear and Eero have already received them by promising future U.S. manufacturing. Fr. Robert points out the absurdity: even if every router were domestic, all core internet infrastructure runs on Chinese-made equipment anyway. In the UK, the Competition and Markets Authority ordered Google to add more prominent links in its AI Overview search results and give publishers a meaningful opt-out — though Jeff Jarvis and Fr. Robert agree this is suicide for any publisher that exercises the opt-out. The Supreme Court delivered the day's sharpest ruling: 8-1 upholding $104 million in FCC fines against AT&T and Verizon for selling customer location data to bounty hunters and data brokers without warrants. Clarence Thomas was the lone dissenter, prompting predictable commentary. The panel agrees the fine is a rounding error but the precedent matters ^{[1] — Leo Laporte "AT&T & Verizon fined $104M for location data sales: The Supreme Court upheld an FCC ruling 8-1 to fine AT&T and Verizon a combined $104 mil…" 2:33:10 Play} .

The week's most culturally significant story arrives at the end: YouTube has officially overtaken Netflix in global average daily viewing, reaching 99.1 minutes to Netflix's 93 ^{[1] — Leo Laporte "Backrooms movie grossed $100M+ in first weekend: Kane Pixels' Backrooms film, made on a $10 million A24 budget by a 20-year-old director, e…" 2:34:42 Play} . Simultaneously, Backrooms — directed by 20-year-old YouTube creator Kane Pixels with $10 million from A24 — grossed over $100 million in its opening weekend. Jeff Jarvis notes this is categorically different from the failed first wave of YouTube-to-Hollywood transitions a decade ago, when Nickelodeon imported creators and stripped out what made them work. This time, the creators are making movies on their own terms. Jeff connects this to Robert Tercek's appearance on Intelligent Machines discussing AI-generated films premiering at Tribeca. Fr. Robert highlights Viva La Dirt League as a YouTube channel with genuine filmmaking chops, and Joey recommends 'The Archive Inbetween,' a YouTube channel doing AI-generated '1950s civil defense films for the multiverse.' South Korea leads global YouTube viewing at 161.5 minutes a day; France recorded the biggest growth.

Leo closes the ad block with Helix Sleep, sharing his personal experience of replacing an eight-year-old mattress. A Wesper sleep study found that participants switching to Helix mattresses gained an average of 25 additional minutes of deep sleep — a roughly 50% increase — and 39 more minutes of total sleep per night. Leo emphasizes that Helix mattresses are assembled and shipped from Arizona to order, avoiding the months-in-a-container-ship problem of overseas alternatives. The quiz-based matching system is recommended for all buyers. Listeners get a discount by entering 'Twit' after checkout at helixsleep.com/twit.

The show winds down with genuine warmth. Jeff Jarvis explains the modern audiobook recording process — you can't just roll on from mistakes, you have to pick up at a three-word cue so the editor can match your tone seamlessly. He also discloses he refused to read the 'no AI company may use this content' boilerplate at the end of the book, because he disagrees with it philosophically. Fr. Robert mentions he will be celebrating a 4:30 AM Mass at the Chapel of Saint Clementine, a tiny private chapel next to the Tomb of Saint Peter that holds at most 17 people, attended by dignitaries and diplomats. Leo plays a clip of '6/7,' his Claude-generated Afro Man parody that has apparently become an earworm for club members. The panel plugs Club TWiT as the only way to watch tomorrow's WWDC coverage live, thanks sponsors, and Leo closes with the traditional 'another TWiT is in the can.'