The episode opens with two pre-roll sponsor reads. KPMG promotes its Adaptability Index — a data-driven tool for assessing organizational culture, strategy, and partnerships — directing listeners to kpmg.com. Indeed follows, citing its claim that Sponsored Jobs are 95% more likely to result in a hire, and offering a $75 sponsored job credit to listeners via indeed.com/podcast. Neither ad is integrated into the main show.

After Andrew nearly bungles the name of the Xbox 25th anniversary console, the crew introduces themselves and the WWDC-focused episode ahead. Adam launches the 'Did They Even Test This' segment with a relatable grievance: YouTube and YouTube TV have nearly identical icons on Apple TV, differing only by a tiny 'TV' label and light/dark mode treatment — leading to constant accidental mis-taps. David suggests the obvious fix of just having a YouTube tab within the main app; Adam explains that platform-specific deals prevent this on Apple TV. Marques proposes the perfect solution: resurrect the original iOS YouTube app's TV-set icon for YouTube TV, leaning into nostalgia.

David announces that a fan created a dedicated 'Did They Even Test This' subreddit, inviting listeners to submit their own examples. The conversation pivots to the WWDC keynote, where Marques, Andrew, and David compare production notes: this year's presentation featured notably handheld camera work, people casually walking in the background of Apple Park shots, and visibly natural broad-daylight lighting instead of Apple's usual ultra-diffused, hyper-polished studio look. No drone shots appeared during the macOS name reveal — a first. Rufus delivers what the group agrees is the most plausible explanation: Apple intentionally made the production feel human and naturalistic precisely because they were spending over an hour selling AI, wanting to counterbalance any sterile or robotic association.

With David having attended the event in person, he fills in what millions of livestream viewers never saw: before the broadcast began, Craig Federighi walked out to a packed Apple Park audience and acknowledged Tim Cook's 15 years of leadership to a standing ovation. Cook followed with approximately 5 minutes of heartfelt remarks calling the tenure 'the honor of my life,' thanking the crowd repeatedly. The crew debates why Apple kept this off the official stream — Adam suggests Cook himself would want the focus on the product and the developers, not on himself. The group notes that Cook remains CEO until September 1st, when John Ternus takes over, meaning the September hardware event will be Ternus's debut.

Marques frames iOS 27 as Apple's Snow Leopard sequel — referencing the 2009 macOS update that shipped with 'zero new features' and was celebrated for making everything faster and more reliable. David runs through the headline performance numbers: apps launch up to 30% faster, AirDrop is up to 80% quicker, Wi-Fi-to-cellular switching is faster and smarter, and the CPU scheduler improvements benefit older devices all the way back to the iPhone 11 — a 7-year-old phone. The crew unanimously celebrates the philosophy. Marques singles out the long-overdue fix for AirDrop's notorious shifting UI that causes users to tap the wrong contact. The hosts also note that Apple specifically improved the corner radii consistency across icons and apps, something power users had complained about for months after the Liquid Glass redesign.

One of the most-mocked design decisions in iOS 26 was Liquid Glass's extreme translucency, which made text hard to read and spawned countless Twitter mockups demanding a simple opacity slider. iOS 27 delivers: a slider lets users choose between ultra-clear glass, a middle ground, and a more opaque option. David clarifies from hands-on experience that the middle position isn't more transparent than either of the original shipped versions — it sits between them. More importantly, Apple applied universal edge-sharpening to all Liquid Glass elements, giving the translucent UI a crisper perimeter that makes content easier to distinguish. The crew notes this is the rare case where Apple explicitly responded to public complaints with a direct UX solution.

iOS 27's Photos app gains three new editing tools that use on-device generative AI. Cleanup — Apple's object-removal tool — is meaningfully improved for reconstruction quality. Extend lets users generatively add up to 25% more image content to any edge of a photo, useful for reframing shots for Instagram's square format or recovering cut-off limbs. Each pass is limited to 25% but can be applied repeatedly. The star is Spatial Reframe: using Gaussian splatting technology borrowed from Vision Pro, it reconstructs any photo as a 3D scene and lets users reposition the virtual camera — rotating perspective, adjusting elevation, and generating new background content that didn't exist in the original shot. David tested it on a photo of The Verge's Vi Song and successfully added arm pixels and draping sweater fabric that looked convincingly real. Rufus found a higher-resolution Fujifilm X100 photo produced even better results than an iPhone photo, possibly due to sharpness advantages.

Apple's Genmoji feature — which generates custom emoji from user inputs — was rebuilt with a new interface that accepts any photo as a base rather than requiring a preset person or emoji combination. The new version also allows mixing custom Genmoji with real Apple emoji. David demos it live on screen by attempting to create a Zelda-themed Link version of his face, with results the crew agrees are 'not that bad' while noting the AI-generated white outline artifact. Image Playground, Apple's standalone AI image generator, also gets photorealistic generation and new editing tools, prompting Marques to quip that the 'slop generator' just got significantly more capable — including generating personalized wallpapers.

The headliner for iOS 27 system features is the Passwords app's new agentic capability: with user permission, it will log into websites one by one, change compromised or weak passwords to 30-character alphanumeric strings, then save the new credentials. Marques calls it polarizing — great for security, but a lock-in mechanic for the Apple ecosystem. The crew debates failure modes: what happens with CAPTCHAs? David notes it needs email access to receive one-time verification codes, and Marques adds that it can be exported to third-party managers like 1Password. Apple also adds physical-to-digital pass scanning in Wallet for NFC-capable cards. The receipt-splitting feature uses Siri's camera to itemize a restaurant bill, assign items to individuals, calculate tip percentages, and request payment via Apple Cash — though the crew agrees this is largely a tech demo since most people just Venmo each other anyway, and immediately Sherlocks the existing app Splitwise.

The hosts sprint through iOS 27's many 'Snow Leopard' fixes and smart additions. Shortcuts now lets users describe what they want in plain language and builds the Shortcut automatically in the editor — Marques immediately approves it as 'good AI.' Independent volume controls for alarms, timers, and media finally arrive, solving the perennial iPhone problem of sleeping through an alarm because you turned the volume down for a late-night video. Find My now supports custom location-sharing durations rather than the previous 1-hour, end-of-day, or indefinite options. Natural-language calendar event creation is declared Sherlock #2 — a direct hit on Fantastical. The Wallet app gains a feature to scan physical passes and digitize them, including NFC card capabilities. Camera UI is redesigned with three accessible settings near the shutter button, and a new Siri mode for visual intelligence replaces the previous Camera Control workaround.

MacOS gets a new name — Golden Gate — and with it a set of thoughtful productivity improvements. The biggest for power users: the Mac finally remembers which apps were on which monitor when you unplug and replug displays, a pain point Windows solved years ago. Sidebars now extend to the screen edge, making better use of available real estate. Safari gains the ability to generate custom browser extensions from plain-English descriptions, surfacing App Store alternatives first if they exist but building from scratch on demand — a move the crew sees as potentially reviving Safari's developer ecosystem. AI-powered tab grouping can organize open tabs by semantic topic, though the logic around ambiguous content (a Google Doc about cats near a pet store tab) remains unclear. Safari also adds a webpage-change monitoring feature — check once daily for updates — which the crew agrees is too slow for concert ticket scalping but useful for low-stakes monitoring.

iPadOS 27 inherits nearly all iPhone improvements plus one iPad-specific highlight: file transfers to and from external drives are up to 5 times faster. WatchOS 27's support list generated controversy: Apple initially announced Series 10 and up, corrected it to Series 9 and up, but either way the Apple Watch Ultra 1 — a relatively recent premium product — is excluded, surprising the crew. Vision Pro gains a floating Siri orb (Adam's favorite announcement), the ability to use personal panoramas as immersive environments (David offers 60+ on his website), real-world object identification via Siri, and a new Iceland environment. The HomePod revelation lands as the sharpest irony of the update: Apple's dedicated Siri smart speaker is the only Apple product that won't receive the new Siri AI, because it simply lacks the RAM to run any of the new models — leaving hundreds of dollars of hardware on the old, limited assistant.

AirPods finally gain a user-adjustable EQ in iOS 27 — high, mid, and low bands, nothing more. Rufus, visibly relieved, immediately declares it overdue by at least a decade. He then goes into granular detail on why he still prefers his AirPods Pro 2 over the Pro 3: a resonance around 600–2000 Hz makes keyboard clicks sound thocky, closing a car door sounds like an explosion, and the noise cancellation is so aggressive it creates discomfort. The 3-band EQ won't solve those specific issues, but Marques notes that for the first time in AirPods history, he can now truthfully say 'you can make them sound the way you want' in reviews. Adam raises the apparent contradiction with Apple's 'adaptive EQ' marketing, which Marques explains operates on a different layer than the user-facing slider.

The episode's mid-show ad block features three integrated reads. Marques delivers the Zapier spot, highlighting the platform's ability to connect AI models like ChatGPT and Claude to existing business tools and noting that teams have automated over 300 million AI tasks on the platform. The Framer read positions it as an enterprise no-code web builder used by companies like Perplexity and Miro, offering 30% off a Pro annual plan at framer.com/wave. Andrew and Marques co-deliver the ChefIQ spot, promoting the Chef IQ Sense smart meat thermometer for Father's Day with 40% off using code WAVE at chefiq.com.

Andrew, a new parent, walks through iOS 27's most significant parental control changes: a redesigned child account setup assistant with preset bundles ('essentials only' locks to FaceTime, Phone, Messages, Maps, and Settings), the ability to convert existing accounts to child accounts, gradual app unlocking as kids earn trust, and request-based access to new apps and websites that sends notifications to parents for approval. Screen time can now be restricted by time of day — during school hours, for instance — and parents can adjust allowances on the fly as rewards or consequences. Marques cuts through the feature showcase with a cynical but accurate observation: the real beneficiary is Apple's future market share. Get a kid comfortable on iPhone early and they are statistically far more likely to remain an iPhone user for life. The crew acknowledges this doesn't make the features bad — just commercially motivated.

Marques sets up the new Siri's core value proposition: unlike ChatGPT or Gemini, Siri's edge is deep integration with your personal data. It indexes your iMessages, email, calendar, and notes to answer questions like 'when does mom's flight land?' by searching your own messages. The three announced pillars are personal context, on-screen awareness (essentially Circle to Search brought to iPhone), and in-app agentic actions that can chain tasks across apps. Marques raises his biggest question: does any of this work with third-party apps? David tests it live by asking Siri to send a Gmail to a colleague — Siri composes the email in Apple Mail instead. Even specifying Gmail explicitly still routes to Apple Mail. David had similar luck adding WWDC briefing events to his calendar: Siri pulled the info from email and added events in chunks, but only to Apple Calendar, not Google Calendar. The conclusion is clear: Siri AI is designed around Apple's first-party app stack.

David breaks down the architecture behind Siri AI's five foundation models: AFM 3 Core (3B parameters, handles simple on-device tasks like timers), AFM 3 Core Advanced (20B parameters using mixture-of-experts, restricted to iPhone Pro and Air with 12GB RAM), plus three private cloud compute models for heavier workloads including AFM 3 Cloud Pro, which specifically uses Google Cloud infrastructure and NVIDIA GPUs. The most surprising detail: Apple is not deploying Gemini directly. Instead, it licenses a 1.2 trillion parameter Gemini model for approximately $1 billion per year to 'teach' its own foundation models — a distillation strategy. Marques unpacks the analogy: if a car company hired a rival to teach them how to build cars using their own materials, the resulting vehicle would walk and talk like the teacher's car but be technically proprietary. Meanwhile, Google pays Apple ~$19 billion per year to be the default search engine, creating an extraordinary mutual financial dependency at the heart of the AI race.

The crew discusses Siri's new interaction design: it extends from the Dynamic Island as a persistent overlay on iPhone, always on top of whatever you're doing, mirroring how Gemini operates on Android. A dedicated Siri app acts as a memory bank for past queries and syncs across all Apple devices, picking up context between sessions. On the Mac, Siri is built into Spotlight (Command+Space) and can trigger on any question or action; a Siri option also appears in the screenshot tool for Circle-to-Search-style on-screen actions. Marques raises one pointed UX critique: unlike Google's assistant, Siri doesn't transcribe your words in real time as you speak — it just shows an orb, then delivers the text output. This makes it harder to course-correct mid-sentence. David's burrito-restaurant test showed Siri could handle complex multi-parameter queries and return mostly relevant results. When David asked about his flight's boarding time and Siri couldn't find it, instead of saying 'I can't help with that,' it offered the general rule that most flights board 40 minutes before departure — a meaningful improvement over the old Siri's dead-end responses.

David and Marques deliver the BetterHelp ad, citing the company's 2026 State of Stigma report: 85% of Americans say getting mental health support is smart, but 74% say society discourages it, and more than 3 in 4 Americans reported anxiety or depression symptoms in the past two weeks. They direct listeners to betterhelp.com/voxpods. The Pure Leaf spot follows, introducing Caffeine Focused — a sugar-free, calorie-free sparkling iced tea using naturally occurring caffeine from black tea and added L-theanine, available in peach and raspberry flavors.

Andrew revisits the Xbox Series X|S 25 anniversary edition in transparent green, genuinely excited by its nostalgic design reference to the original translucent Xbox. Marques admits he's never seen the original console it references, but likes the look anyway. David pivots to his main grievance: Nintendo's Ocarina of Time remaster, which a trailer showed heading toward hyper-realism. David's argument is architectural: cartoon and cel-shaded art styles age gracefully because they were never trying to mimic reality. Hyper-realistic graphics always become dated. He cites Wind Waker as proof — it looks as fresh today as it did two decades ago. Hyper-real Ocarina will not look the same in 10 years. He reluctantly reserves final judgment for gameplay footage while still stating he 'massacred my boy.' Marques offers a market-logic counterpoint: if you don't buy it, you send a signal. If you do, you fund more remasters.

David returns to a recurring podcast topic: Anthropic's Claude Fable, previously withheld as too dangerous to release, is now available to all paid subscribers — but only for one month before being restricted to paid API access. Rufus explains Anthropic's stated rationale: exposing the model broadly allows them to observe jailbreak patterns at scale, which trains their alignment model to be more robust. David remains skeptical, calling it 'the drug dealer playbook': give a free taste, then charge per gram. The economics back this up — $20/month AI plans cost companies approximately $5,000 per user, making API-per-token pricing essential for profitability. On the same WWDC day, OpenAI filed 'confidentially' for an IPO, a filing that immediately became public news — prompting Andrew to wonder if tech companies understand what 'confidential' means.

David briefly notes one understated development: Apple and ChatGPT appear to be pulling apart. The previous Siri behavior of asking users 'mind if I ask ChatGPT this?' is no longer necessary now that Siri itself has robust language model capabilities. The crew then pivots to the 'Were You Paying Attention' trivia round, hosted by Adam. Question one asks for the official color of the Golden Gate Bridge — Marques correctly answers 'international orange,' connecting it to the Apple Watch Ultra's signature button color. Andrew humorously writes 'Nixon Five' as his answer. Question two asks where the Vision Pro backpack demo flight was heading — the answer is Iceland, which David had the Vision Pro environment connection but still missed. Adam tallies scores (Marques 26, Andrew 25, David 31 before this round) before Marques closes the show, inviting listeners to leave requests in the comments for specific Siri AI tests they want to see in future reviews. The episode ends with the customary Kool-Aid sign-off.