Speaker
Joey de Villa
Appearances over time
1 episodes
Episodes
1
Podcasts
Quotes & moments
SpaceX's planned IPO would value it at $1.77 trillion — three times Saudi Aramco — and yet analysts estimate only 54% of promised data center capacity will be delivered by 2030. The panel tears apart the thesis that this buildout makes economic sense, noting that when old data centers go dark they're useless, unlike dark fiber which could be relit for pennies.
Anthropic says more than 80% of code merged into its codebase is now written by Claude, up from single digits in early 2025. Claude's success rate on open-ended problems jumped from under 20% to over 70% in months. The panel debates whether this is genuine recursive self-improvement or a carefully timed IPO narrative.
Fr. Robert Ballecer invokes Charlie Munger's maxim — 'show me your incentives and I'll show you your outcomes' — to explain why calls for an AI pause are performative. The race to build the first fully functional AGI has no second-place prize, making any genuine pause structurally impossible regardless of rhetoric.
Fr. Robert Ballecer, speaking from direct experience, says Anthropic's Mythos model performs intelligent security analysis in minutes that used to take experienced professionals days or weeks. It's not a script kiddie tool — it reasons from a vast knowledge corpus to find genuine software vulnerabilities.
Leo Laporte draws an analogy from Shannon's theorem: just as digital audio can represent analog sound indistinguishably if sampled finely enough, digital AI may represent understanding indistinguishably from the real thing. Fr. Robert Ballecer counters that true consciousness lies in accumulated variance — the billionth firing of a human brain looks nothing like the first.
The Vatican developed a completely private AI model trained from scratch on several hundred years of Church documents. It speaks Latin, handles 20-language real-time translation, and is used for sensitive canonical cases that can never legally touch the internet. Spanish regional accents are proving the hardest to crack.
The panel reaches consensus that AI's greatest value lies in hyper-specialized deployment, not in building one giant general-purpose model. The Vatican's AI never hallucinates because it simply says 'I don't have enough information' when probability drops below threshold — a feature only possible when scope is deliberately constrained.
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 that at $10,000 for dual SPARK hardware, you're buying years of frontier model tokens. The real value of local AI is privacy — especially for the Vatican, which legally cannot let sensitive documents touch the internet.
A 20-year-old YouTube creator got $10 million from A24 and made over $100 million at the box office in one weekend. Meanwhile YouTube globally overtook Netflix in daily average viewing for the first time. The panel argues this isn't a fluke — it's the permanent collapse of traditional gatekeeping in entertainment.
The Financial Times reported OpenAI plans to transform ChatGPT into a super app, promoting its Codex coding product and shifting from chatbots to AI agents. One senior employee said 'chat is dead' and argued the product needs renaming. The panel is skeptical this will work — the history of super apps is littered with failures.
Trump said on Air Force One that government ownership of AI companies could be 'a beautiful thing' that makes Americans rich. Sam Altman had already floated the idea. But the panel notes the U.S. government already took a 10% stake in Intel via the CHIPS Act — so the precedent exists, even if the implications are alarming.
Major homebuilder Pulte claims it can deploy 8,000 backyard mini data centers 6 times faster at 5 times lower cost than a 100-megawatt facility. Fr. Robert Ballecer explains why this fails on bandwidth physics alone: NVIDIA's Vera Rubin architecture uses 1.6 petabytes per second of interconnect — roughly 160,000 times faster than residential fiber.
TWiT's network was breached in January 2026, and the intruder stayed undetected for 121 days — more than 4 months — because Leo had disconnected the Thinkst Canary for an ad read. The breach was only discovered when Google flagged unusual activity in TWiT's Workspace account.
Wired discovered face-recognition code called 'Name Tag' silently embedded across millions of phones in Meta's smart glasses AI app, pulling biometric face prints from Meta's servers. Meta hasn't enabled it, but the capability is built and waiting. Fr. Robert Ballecer has spent 7 years poisoning facial recognition databases with fake tagged photos to fuzz his identity.
Analysis
What they talk about
- Business 25%
- Government 25%
- Technology 25%
- TV & Film 25%
Connections
Shows they appear on and people they share episodes with. Drag to explore.
