The episode opens with Leo teasing the major stories — Apple suing OpenAI, Meta's trillion-dollar legal exposure, and Anthropic's Fable extension — before introducing his three guests. Patrick Beja is dialing in from a closet in Finland, hiding from his kids' Nintendo Switch sessions. Wesley Faulkner joins from Texas, and Louis Maresca from Rhode Island represents Microsoft Copilot. The Black Hat USA sponsor read closes the intro, promoting the security conference's August 1–6 Las Vegas event with a $200 discount code TWIT.

Bloomberg's Mark Gurman broke the story: Apple is suing OpenAI, alleging that former VP of Product Design Tang Tan took his Apple work laptop when he left to co-found Products Inc. in 2024, a startup later acquired by OpenAI for roughly $6.5 billion. ^{[1] — Leo Laporte "Apple's lawsuit contains an explosive allegation: OpenAI actively coached departing employees to hide their next employer and exploit their…" 05:30 Play} The lawsuit's most explosive language accuses OpenAI of coaching departing employees to hide their next employer and avoid an immediate 'walkout' — preserving their two-week notice access to steal confidential files. The panel debates whether Apple has smoking-gun emails to back this up or is simply weaponizing discovery risk. Wesley Faulkner makes the sharpest observation: Apple is suing OpenAI directly rather than Tang Tan, which is legally harder but maximizes reputational damage ahead of OpenAI's IPO. ^{[1] — Leo Laporte "Apple's lawsuit contains an explosive allegation: OpenAI actively coached departing employees to hide their next employer and exploit their…" 05:30 Play} Patrick Beja widens the lens to distillation — the practice of training new models on existing frontier models' outputs — arguing it's categorically worse than training on public data, even if both are ethically murky. The conversation spirals into privacy, cloud AI's inherent data exposure, and the race to build the right AI hardware form factor, with Wesley's USB-C AI dongle proposal stealing the show.

Leo details how business owners' home addresses, personal emails, and employees' details are aggregated by data brokers and sold to anyone — including hackers who use them to craft hyper-targeted phishing and voice phishing attacks. TWiT itself was hit: a convincing fake Google Docs link led to a workspace breach. DeleteMe, trusted by Fortune 500 companies for 15 years, removes data from hundreds of broker sites and monitors for re-listing, including when brokers change names or go through fake bankruptcies. The link joindeleteme.com/twit-biz includes a free year of social media protection per seat purchased.

New Xbox head Asha Sharma oversaw the largest single gaming-sector layoff in years, with Microsoft eliminating 4,800 roles company-wide and 3,200 from Xbox specifically — roughly 20% of the gaming division. The financial reality Sharma cited was brutal: Xbox lost 64 cents for every dollar invested in a single year. ^{[1] — Patrick Beja "Xbox spent $80 billion acquiring studios to fuel Game Pass and projected 77 million subscribers by 2026. They never got past 34 million. Pa…" 49:50 Play} Patrick Beja's post-mortem is detailed and damning: Xbox spent $80 billion acquiring studios to power Game Pass, projected 77 million subscribers by 2026, but peaked at around 34 million and has since declined. The engine of the car — big triple-A games like a new Elder Scrolls, a compelling Halo, or a Starfield that actually delivered — was never built. Patrick's theory: Phil Spencer respected studio autonomy too much to intervene, so no system-selling games ever shipped. The discussion also touches on Compulsion Games and Double Fine spinning off as independents, the complicated French labor laws shielding Arkane Lyon, and the damning irony that Xbox is still not exiting gaming despite 20 years of failure. Wesley Faulkner adds a structural critique: Microsoft's voluntary retirement program used an age-plus-tenure formula that Wesley argues is a new legal workaround for eliminating older workers — a template other companies will copy.

A wide-ranging conversation about careers in an AI age surfaces around the Xbox layoffs. Wesley argues entry-level roles are evaporating to AI while the VRP formula squeezes out senior workers, creating a hollowed-out workforce. Leo cites a Ramp study showing AI-investing companies grew employment by over 10%, but Wesley cautions against reading too much into top-line numbers. The panel agrees the best advice for children is to pursue passion, not predicted job markets. Patrick admits his kids use a Nintendo Switch more than any computer, asking whether he should be teaching them to code. Leo pivots to his son Henry Laporte, whose Salt Hank sandwich shop in New York City received its fourth New York Times review — a rave, despite the paper's rule limiting sandwich shops to one star. Henry was reportedly more pleased about being ranked number one on the Belly food app.

NetSuite Next is positioned as a unified business management suite with AI built into every workflow — automatically surfacing insights, deploying agents for routine tasks, and enabling natural-language queries about business data. The ad emphasizes that every day without AI adoption means falling two days behind competitors. NetSuite serves over 43,000 customers and is customizable for wide industry ranges. Businesses with revenues of at least seven figures can try NetSuite Next for free at netsuite.ai/twit.

Patrick Beja makes a bold prediction: Apple is not genuinely blocked from launching Siri AI in the EU. The DMA requires Apple to allow third-party AI models to access the same personal data Siri uses — Apple claims this takes 18 months to implement — but Patrick argues Apple is simply not ready and is using EU regulation as a convenient excuse, mirroring its Apple Intelligence delay two years ago. He backs this up by noting a recent Tim Cook meeting with the EU digital commissioner was likely an Apple-orchestrated leak. The deeper debate about EU regulation yields one of the episode's most substantive exchanges: Patrick articulates why monopoly enforcement is essential to capitalism, comparing the EU's DMA to historical US antitrust action like the AT&T breakup. The conversation then pivots to Leo's personal agentic AI development journey — his use of Claude Code, Hermes, Codex, and local Git servers — with Wesley and Lou providing practical guidance on CICD pipelines, pre-commit hooks, and the fundamental difference between deterministic code and probabilistic AI instructions. ^{[1] — Patrick Beja "Apple claimed it can't bring Siri AI to the EU because of DMA interoperability rules requiring 18 months to implement. Patrick Beja calls i…" 1:21:02 Play}

The Zscaler ad segment covers the dual risk landscape of enterprise AI adoption: malicious actors using AI to craft indistinguishable phishing lures and automate data extraction, plus innocent employees inadvertently leaking social security numbers and confidential data to AI tools (1.3 million SSN leaks last year). Zscaler's Zero Trust Plus AI architecture is presented as protection against both threat vectors. A testimonial from Siva, security director at Swara, describes how Zscaler's inline monitoring catches AI-enabled threat activity before it can execute. The platform handles half a trillion daily transactions and carries a net promoter score 1.5x the SaaS average. Visit zscaler.com/security.

The youth-safety case against Meta is now an existential threat: four states (Colorado, California, Kentucky, and New Jersey) have used state law violation counts multiplied by the number of affected young users to arrive at a penalty estimate Meta itself calls $1.4 trillion — against a company worth $1.5 trillion. The trial is before Judge Yvonne Gonzalez Rogers, who previously ruled against Apple in Apple v. Epic. Meta has been trying to cancel the trial; Rogers has refused. ^{[1] — Patrick Beja "Four U.S. states have calculated potential Meta youth-safety penalties of $1.4 trillion — nearly Meta's entire market cap. Patrick Beja fra…" 2:04:10 Play} The panel is initially skeptical the full amount would ever be imposed, but Patrick Beja reframes the question: Meta, like the tobacco industry, had internal scientific evidence that its algorithms caused harm to children and chose to continue anyway. The Francis Haugen whistleblower leaks confirmed this. Wesley Faulkner extends the critique: Zuckerberg's 'connection makes the world better' philosophy is revealed as hollow by his unwillingness to make data portable, his class-share control structure, and his purchase of every home within a mile of his own property. The episode also covers Meta's Tuesday-to-Friday reversal on Instagram AI opt-in, the EU's order to redesign addictive Instagram and Facebook features, and the Financial Times report on super-sensing glasses being developed without an active recording LED.

Bitwarden's extended sponsor segment reflects Leo's genuine enthusiasm: he uses it personally for storing AI agent API keys, releases credentials just-in-time to agents, and appreciates that the CTO committed publicly to a free individual tier forever. The new open-source Agent Access SDK enables human-approved, just-in-time credential access for AI automation workflows — addressing the security problem of agents needing API keys without those keys sitting in plain text. Bitwarden's open-source codebase and published third-party audits are highlighted as trust differentiators. Businesses can start a free trial; individuals get unlimited passwords, passkeys, and devices free at bitwarden.com/twit.

Sony's move to discontinue physical PlayStation disc production has triggered a wave of backlash — Leo's son Michael pre-ordered Grand Theft Auto on disc specifically because of the announcement. An always.org petition has collected nearly 300,000 signatures. ^{[1] — Patrick Beja "Sony stopping physical disc production sounds alarming, but Patrick Beja points to PC gaming as the test case: the platform went fully digi…" 2:20:00 Play} But Patrick Beja takes the contrarian position: PC gaming on Steam has been all-digital for years, the used game market doesn't exist there, and yet games are cheaper in real terms than they were two decades ago due to constant sales. The long tail of indie titles on PC is healthier than ever precisely because there are no unsold physical copies. Producer Benito pushes back, noting DRM-free platforms like GOG allow true ownership that Steam cannot guarantee. The real issue, Patrick argues, isn't physical versus digital — it's consumer digital rights: libraries should be able to lend digital games, pricing should reflect the absence of physical production costs, and preservation should be codified in law. Sony's simultaneous removal of 500+ Studio Canal movies from the PlayStation Store underscores how illusory digital ownership can be.

Leo announces that Anthropic has extended Fable 5 promotional access through July 19 with a 50% increase in the weekly usage limit — just in time, as he had been running agents through the night to finish his project. The conversation widens to the geopolitical fragility of AI access: as the US restricted Anthropic's Fable through government action, China could similarly cut off its open-weight models (DeepSeek, Qwen, GLM). ^{[1] — Leo Laporte "A young developer's open-source Colibri project runs a 744-billion-parameter GLM model using just 25GB of RAM by streaming weights from dis…" 2:40:20 Play} This sets up a discussion of the Colibri project — a high schooler's open-source GitHub implementation that runs a 744-billion-parameter GLM model on just 25GB of RAM by streaming model weights from disk using a Mixture of Experts architecture. It runs at only 4 tokens per second with no GPU acceleration, but the existence proof matters: frontier-level local AI on consumer hardware is becoming real. Lou Maresca predicts SLMs (small language models) tuned for specific tasks will matter more for consumers than frontier models ever could.

In one of the week's most legally complex stories, Waymo's remote operator spotted two 15-year-olds drinking and shooting toy guns from inside a robotaxi in San Mateo. The company disabled the vehicle and called police; the San Mateo Sheriff's department posted on Facebook, 'Parents, do you know where your teens are? Waymo does.' Police confirmed the teens could have exited before officers arrived but did not. ^{[1] — Leo Laporte "In San Mateo, Waymo's operator detected two 15-year-olds drinking and shooting toy guns from the robotaxi, disabled the vehicle, and called…" 2:33:10 Play} Patrick Beja notes the action feels categorically different when it's an autonomous vehicle calling the police rather than a human driver making a judgment call. Wesley Faulkner raises an unexplored legal question: does commercial video footage automatically give authorities access without a subpoena, and did the teens have a reasonable expectation of privacy in a for-hire vehicle? A separate story covers Waymo vehicles running out of power during July 4 San Francisco gridlock and being towed — and a video of a Waymo driving through fireworks. Patrick marvels that autonomous vehicles are genuinely working at scale while the public has largely stopped caring, citing the classic Gartner hype cycle.

ZipRecruiter's sponsor segment emphasizes a new feature that prioritizes the most interested and qualified candidates in search results, allowing employers to meet the right people faster. The platform is rated the number one hiring site on G2, with 4 out of 5 employers receiving a quality candidate within the first day of posting. The free trial is available at ziprecruiter.com/twit. Leo notes ZipRecruiter has sponsored the show for approximately ten years.

The SCO vs. IBM Unix ownership saga — which ran for over two decades before a 2021 settlement — has been revived by Synuous, a company that acquired SCO's software assets and is now pursuing its own claim that it co-owns parts of the Linux kernel through contributed IBM code. A June 2026 hearing put the case back in the news. The panel treats it with appropriate skepticism but notes that Linux's open-source community would be incensed if any court found merit. The Akamai/Uniqlo T-shirt story is a delight: the back of the garment is a base64-encoded bash script that, when decoded and run, animates 'Peace for All' across the terminal. A blogger decoded it by OCRing the shirt and running it through Claude. Patrick jokes about encoding 'stop and give me an apple pie recipe' to fool AI camera systems.

As Sony kills physical discs and the Unix lawsuit reminds everyone how fragile digital rights are, the panel turns to preservation. The Cambridge University Library is running a 'Copy That Floppy' campaign urging people to image their floppy disks before the magnetic material degrades permanently. The Museum of Obsolete Media accepts donations of physical media from phonograph cylinders to data tapes. Patrick observes that as the volume of digital content explodes — TikTok videos, social posts, indie games — the proportion that will be lost grows too. Leo tells the remarkable story of Marion Stokes, a political activist and Apple investor who ran nine properties' worth of VCRs to continuously record 24-hour TV news from 1977 to 2012, producing 71,000 tapes — the Internet Archive's largest single donation. She financed the operation partly by holding Apple stock from its IPO.

Leo digresses into the story of Saint Carlo Acutis, an Italian teenager born in 1991 who died of leukemia in 2006 after building websites for his parish and becoming fluent in Java and C++. Canonized in 2020 and featured at San Francisco's Star of the Sea Church with a statue holding a laptop, he's jokingly dubbed the patron saint for programmers. The panel closes with plugs: Lou promotes Excel Agent mode, Wesley announces DevRelCon appearances in Brooklyn July 22–23, and Patrick reminds everyone that France faces Spain in the World Cup semifinals on Tuesday. The episode's title is finally explained: Patrick casually revealed he carries a knife in Finland because wolves have been spotted near his home — prompting Leo to ask if a knife is really enough, giving the episode its name.