The episode opens with two back-to-back pre-roll sponsor spots. Indeed promotes its Sponsored Jobs product, claiming a 95% higher likelihood of hire versus non-sponsored listings and offering listeners a $75 job credit at indeed.com/podcast. BetterHelp follows with a mental health pitch anchored in its 2026 State of Stigma report, which found that 74% of Americans believe society discourages asking for help — a data point designed to lower the barrier for trying online therapy.

Before the formal intro, the hosts fall into a spontaneous game of naming every macOS version named after a California location — running from Mavericks and Yosemite through El Capitan, Sierra, Mojave, Catalina, Big Sur, Monterey, Ventura, and Sonoma. The riff is light and funny, but it plants a seed for a later conversation about Apple running out of recognizable California place names and the leaked 'Big Bear' moniker for the next macOS release.

Marques formally welcomes listeners to Episode 559 and runs through the agenda: it's the week before WWDC, so predictions are on the table; Andrew has a new DIY watch to show off; NVIDIA's RTX Spark is promising Apple Silicon competition; and there's a bundle of Microsoft Build news to work through. The hosts also nudge viewers to subscribe on YouTube to make sure notifications actually fire — setting up a running joke about chapters breaking.

Ellis has been trying to broaden his horizons beyond Apple products, cycling through Garmin watches and now attempting to log PT exercises in Garmin Connect. The app's exercise search is the target of his 'Did They Even Test This?' — typing 'leg' returns nothing, 'press' returns everything, and 'single leg leg press' is simply unfindable. The hosts explore whether the problem is Ellis's beginner status or genuinely broken keyword logic, landing on both. The conversation drifts to whether macro fitness data is even useful for beginners and whether Garmin will add AI coaching behind a subscription. The segment ends with Ellis requesting crowdsourced help from listeners.

The hosts trade opposite anecdotes about the Fitbit Air's cultural footprint: Ellis's non-techie friend was talked out of buying one by the Waveform podcast, while Marques encountered four or five people at California airports proudly showing off Fitbits they bought because of him. David synthesizes the paradox — citing Christian Selig's identical conclusion — that the Fitbit Air is a product you can simultaneously love, enjoy daily, and honestly describe as having a terrible app. Adam adds that as long as the basics work, the AI coaching is a nice-to-have that doesn't make or break the product.

A social media trend of attaching traditional watches to the Fitbit Air strap — popularized partly by a photo from Dan Seifert, now at Google — has convinced Google to release official Fitbit Air hardware specs. This opens the door for third-party makers to build adapters, straps, and even bicep attachments. The hosts geek out on the potential: Andrew wants notifications from it, Adam is already dual-wristing without drama, and David imagines a future where swappable Fitbit Air bands cost under $30 each. The segment reveals a fundamental tension between minimalist wearables and feature creep that the hosts can't quite resolve.

The conversation turns technical when the hosts investigate whether flipping the Fitbit Air to the underside of the wrist actually works for fitness tracking. Marques pulls up an AI overview to explain the mechanics: the optical sensor needs a tight light seal against the skin, and the tendons running from the forearm to the fingers constantly flex and introduce light leak. Adam notes that bicep straps gain popularity precisely because muscle tissue provides a cleaner optical read. David recalls Des Fitz's review noting that the Fitbit Air's thin profile already risks light leak at its edges compared to the wider Whoop.

Andrew discovered a YouTube video about the Ollee Watch replacement board — a $50 circuit board made by a single Canadian developer that slots into a $25 Casio F91W and adds Bluetooth connectivity. The mod keeps the original screen and buttons but adds an accelerometer for step counting, automatic phone time sync, configurable watch faces, and a phone app for logging timed sessions. It even supports blackjack. The creator ships every order on Fridays no matter when you order, underscoring just how small the operation is. Andrew loves it for the tinkering experience and sees potential if notifications arrive in Q2, though David debates whether the Casio board adds anything the Fitbit Air doesn't already cover.

Andrew makes a direct plea: the Fitbit Air already has a vibration motor, so refusing to add optional notifications is simply withholding a feature for no reason. David and Adam push back — a screenless device exists precisely to let you forget about your phone, and adding vibrations defeats that purpose while killing battery life. Marques draws an analogy to the minimalist phone movement: people buy it intentionally limited, even if logically you could 'just use a regular phone and use it less.' Andrew's final position — default it off, let users opt in — earns a reluctant partial concession, but the philosophical divide stands.

David announces that Google is rolling out a Search Console opt-out toggle for AI overviews, AI mode, and Discover — starting in the UK first, a nod to CMA pressure. Google promises the toggle won't affect regular search rankings but won't stop content from appearing in Gemini. Marques frames the key limitation: training data already in the model is like a black hole — the mass is in, and you can't extract individual pieces. The hosts are curious whether large publishers like Wikipedia will opt out and whether Google will negotiate private deals as it did with Reddit. The segment closes with Marques noting that this is Google's chance to actually prove its claim that AI-driven traffic is higher quality than raw volume traffic.

Marques reads a Shopify sponsor spot, framing the platform as the antidote to the overwhelming to-do list of starting a business. The ad highlights Shopify's ready-to-use templates, built-in marketing tools, and its claim to power nearly 10% of all US e-commerce — including established brands like Gymshark and Magic Spoon. Listeners are directed to shopify.com/wavform for a $1-per-month trial.

Two more sponsor reads: Framer positions itself as the pro site builder where AI agents draft pages directly on the live canvas, enabling teams to ship production-ready websites faster. Listeners get 30% off a Framer Pro annual plan at framer.com/wave. KPMG then pitches its Adaptability Index as a data-driven blueprint for building organizational resilience, directing listeners to kpmg.com/us/adaptability.

WWDC 2026 is days away and the tagline 'All systems glow' makes it clear Siri is the centerpiece — though David darkly notes last year's tagline was also 'It's glow time' and Siri never showed up, and the executive who announced the revamp has since left for OpenAI. Marques expects Apple to lean into Siri's unique advantage: deep, privacy-respecting access to on-device data (contacts, iMessages, calendar, health) that third-party apps like ChatGPT and Gemini can't touch. David and Marques contrast this with Gemini Spark's almost unsettling knowledge of your life, illustrating that Google's utility advantage comes directly from its data collection. The Gemini Daily Brief comparison becomes a confessional: David gets rich, actionable reminders; Marques gets a list of emails he already read; Adam gets saved from missing his Spanish class.

Ellis introduces his trivia question: how many social networking services did Google launch before Google+? He immediately pre-empts arguments by ruling out Dodgeball (SMS-based), Picasa Web (social as afterthought), OpenSocial (a protocol, not a platform), and initially — controversially — Google Wave and Google Buzz. The hosts debate Price Is Right vs. closest-Delta scoring rules, argue about whether Google Wave counts as social media, and spend more time adjudicating the question than answering it. The answers are sealed for the end-of-episode reveal.

A mid-roll block runs three sponsor spots: Middy Health targets women in perimenopause with specialist-led virtual care and insurance-eligible visits; Fetch Pet Insurance promotes up to 90% vet bill reimbursement with any US/Canada vet in-network; and Pure Leaf Mental Focus introduces a new naturally caffeinated sparkling tea with added L-theanine, available in peach and raspberry.

NVIDIA made its move against Apple Silicon with the RTX Spark Superchip, a 3-nanometer chip combining 20 CPU cores, 6,144 GPU cores, up to 128GB of unified memory, and 1 petaflop of AI compute into a laptop form factor. Jensen Huang pitched it heavily as an agent platform, claiming CUDA support for everything NVIDIA has ever made — but the hosts note he also said 'there are only a billion people' when there are 8 billion. Marques offers the clearest verdict: the window for how good or bad this could be is enormous. The DGX Spark desktop precedent — which climbed from $3,000 to $6,300 — suggests pricing will be steep. Ellis sees a genuine niche in post-production workflows. David reports the Microsoft Surface Ultra companion laptop looks like a MacBook with extra ports.

The Surface Ultra is Microsoft's flagship laptop companion for the RTX Spark ecosystem: 15-inch mini-LED display, the largest haptic trackpad in Surface history, a full port complement including HDMI and SD card, and a mysterious larger USB-C port that prompted a knowing smile and 'more to share later' from a Microsoft employee — speculation points to a magnetic breakaway charging solution. But the hosts flag a critical problem: there's no Linux support at launch, and the developer audience NVIDIA is targeting heavily skews Linux. Microsoft's response to Linus Torvalds reportedly amounted to 'we have bigger things to worry about right now,' which the hosts reluctantly accept as a reasonable launch-first position.

Microsoft's Build conference also unveiled Project Solara, an OS designed specifically to run AI agents across diverse hardware form factors. The demo showed an Echo-like desk device and — most eye-catchingly — a smart corporate badge with a camera, letting a presenter record a video and ask an agent to edit and post it to social media. Adam immediately thought of the Rabbit R1; the hosts agree the concept is similar. The OS runs on Android, which Ellis finds deliciously ironic given how long Mac users have had to hear 'this great app is Windows-only.' The lingering question: Microsoft has a history of impressive concept hardware that never reaches consumers, and the Agent OS itself is unproven.

In December Microsoft gave employees free Claude subscriptions, and predictably they used Claude instead of Copilot. Now Microsoft is shutting down Claude Code access by June 30 — the last day of its fiscal year — both to cut costs and to stop subsidizing a direct competitor to its own AI products. David frames the larger implication: we've reached the moment where the cost of running AI agents exceeds the cost of human engineers, and companies are speedrunning this realization. Uber recently abandoned its AI coding initiative for similar reasons. David notes that Copilot and GitHub Copilot both integrate Claude under the hood, so employees retain some access — just not the direct subscription.

David explains the complex mechanics: Elon merged XAI into SpaceX because XAI was burning cash, absorbing its losses onto SpaceX's balance sheet. Then SpaceX convinced the NASDAQ 100 to waive the standard one-year seasoning period, meaning within roughly 15 days of the IPO, the stock automatically enters all NASDAQ 100 index funds — and by extension, millions of 401ks. Ellis corrects himself: the combined entity lost $4.3 billion in Q1 2026 alone, not the $800 million he initially estimated. His analogy is devastatingly simple: it's a meme coin pump scheme, except instead of tweeting to convince people to buy, people are literally being forced to buy through their retirement accounts before the stock has time to correct.

David reports that Anthropic confidentially filed its IPO paperwork — then tweeted about it, somewhat undermining the 'confidential' aspect — with valuations estimated just under $1 trillion. The hosts debate whether Anthropic or OpenAI will go public first, with David betting on Anthropic. Ellis surfaces the bubble concern: with 80% of market index weight in AI companies, even index funds no longer provide meaningful diversification. Hank Green's analysis is cited as a useful primer. The segment closes with a quick not-financial-advice disclaimer and a nod to the meme coin energy of the entire AI market.

Trivia resolution time. Ellis adjudicates his question: Dodgeball, Picasa Web, OpenSocial, and Google Wave don't count under his rules, but Google Buzz does after the hosts push back. Final answer: three — Orkut, Google Buzz, and Google Friend Connect — with Orkut the fascinating outlier that only succeeded in Brazil. David, who guessed three, wins the point and pulls ahead to 30 in the running tally. The bonus question about which Google+ feature became a standalone product in 2013 stumps almost everyone: it's Google Hangouts, which Andrew correctly identifies. The hosts reminisce about Hangouts' many deaths and resurrections as a unified messaging app, mourning the superior Allo that got killed while the inferior Duo survived.

Marques closes by inviting listeners who made it to the end to comment 'deep' as a signal of their dedication, and reminds them to subscribe ahead of next week's WWDC live coverage. David reads the production credits — Adam and Ellis produce the show, distributed through Vox Media Podcast Network, with music by 20syl. A post-roll ad for Goldbelly (nationwide food delivery, 20% off with code GIFT) and a brief recorded cold-open clip wrap the episode.