The delightfully messy state of Apple gadgets

The episode opens with Nilay and David reacting to Apple’s new MacBook Pro with an OLED display and touchscreen. They joke about Apple’s name “ProLED,” which is apparently a portmanteau of “pro” and “OLED,” and compare the launch video to the escalating children’s book “If You Give a Mouse a Cookie.” Nilay says the product is not what he imagines when he thinks of a $4,000 professional laptop, but he is going to buy one anyway because his M1 Pro MacBook Pro is beginning to struggle. The new machine is thinner and lighter, has a better display and newer chips, and loses almost a pound in the larger model. They disagree slightly about whether the touchscreen is lazy or the first step in a larger strategy. David thinks Apple may simply have found that the desired display panels were easier or cheaper to obtain with touch built in. Nilay argues that Apple is just adding a direct input method without redesigning macOS, which may be the right choice: touch could be useful for zooming, sliders, timelines, and creative applications, without becoming the primary way people use the computer. They cite Steve Jobs’s old objection that vertical touchscreens cause arm fatigue, but say Apple is not asking users to hold up their arms all day—only to reach out occasionally when touching is more convenient than using a trackpad.

The hosts interpret the touchscreen Mac as evidence that Apple is in a “delightfully messy era.” The company now sells a Mac that accepts touch but runs macOS, an iPhone-like Duo that has a different multitasking model, and an iPad mini that remains surprisingly limited despite sharing a size with the Duo. Apple’s event demonstrations focused on practical creative tasks: zooming in Final Cut Pro, manipulating a Gaussian-splatting 3D model, and touching controls in photo-editing software. The hosts note that these are useful conveniences, but not a new theory of computing, and question whether the touchscreen is worth the higher price. David argues that Apple’s application business is under pressure from people becoming bored with phones, governments challenging the App Store, social-media regulation, and AI agents that could perform tasks through cloud computers or Linux systems without traditional apps. Nilay adds that the iPad is trapped between the iPhone below and the Mac above. They suggest Apple may now be using the Mac and iPhone to absorb the best ideas from the iPad rather than making the iPad the future of computing, as Apple once predicted. A touchscreen MacBook Air would intensify that squeeze. At the same time, Apple appears to be betting on premium hardware: thinner laptops, bright OLED displays, long battery life, ports, and the ability to serve as the physical “shovel” for an AI gold rush. The hosts conclude that Apple’s safest strategy may be to keep making excellent computers rather than inventing another unreliable Mac era full of broken keyboards and too few ports.

The discussion turns to Instagram, which the hosts describe as “going through it.” Meta is pushing longer-form video, preparing a television app, and trying to make Instagram compete more directly with YouTube and TikTok. But David argues that Meta’s larger AI strategy could damage Instagram itself. Instagram is an exceptionally effective advertising platform: people see products, become interested, and eventually buy them. If an AI agent instead finds products, compares prices, and makes purchases on the user’s behalf, the connection between Instagram’s advertisements and the final transaction could be weakened. Meta hopes an agent might discover a product on Instagram and complete the purchase, but the hosts say the agent could just as easily find a cheaper item elsewhere. They also see a hardware problem for Meta. If Mark Zuckerberg wants users to interact through AI agents, smart glasses, or other devices rather than iPhones, Meta has to decide where Instagram fits. It could be on a television, on glasses with displays, or in some future device, but the familiar Instagram experience would change. The current service is also flooded with AI-generated “slop,” including deepfakes and viral memes that reuse songs, likenesses, and comedians’ material. The hosts say Instagram is particularly vulnerable because it is deeply tied to celebrity and creator culture, yet it lacks the kind of likeness-detection protections they associate with YouTube. They argue that Meta could label or filter obviously AI-generated material, but doing so would conflict with the company’s simultaneous effort to give creators more AI-generation tools.

Adam Mosseri’s public appearances are presented as evidence that Instagram knows it has a serious problem but may not know how to solve it. In an interview, Mosseri discusses a fictional AI comedian who steals jokes from real comedians and says originality is likely the principle that should guide the response. The hosts read this as an implicit acknowledgment that Instagram should prioritize original creators and suppress or identify copied AI content. They also note the conflict: Meta is investing heavily in AI, so Instagram cannot easily attack the same AI systems that its parent company wants creators to use. Instagram is testing “Features,” a push toward longer videos, including 16:9 videos as long as 90 minutes, and is expanding television applications. The hosts say the move makes sense because Instagram is often a discovery layer for YouTube, but creators immediately asked how they would be paid. YouTube offers revenue sharing, while Instagram has historically depended more on brand deals and has not offered a comparable straightforward payment system for long-form uploads. The episode describes Instagram as culturally central in categories such as beauty, fashion, cars, and real estate, but says that many creators still make their main income on YouTube. Meta is therefore trying to close the loop: TikTok has pursued music and TikTok Shop, Instagram is pursuing long-form video, and platforms increasingly want users to discover, watch, buy, and transact without leaving. The hosts question whether creators will move their main work unless Instagram offers meaningful revenue sharing.

The hosts connect Instagram’s strategy to a broader “loop-closing” competition among platforms. TikTok is testing a restaurant guide because users already discover restaurants in videos. Reddit is moving toward a more logged-in, personalized For You experience because it wants a direct relationship with users instead of relying on Google referrals. Reddit’s problem is that Google’s AI Overviews are reducing the number of people who click through to Reddit after searching. Reddit reportedly saw search referrals fall by more than one million per week on average during October, with the decline accelerating into the fourth quarter. The hosts say Reddit previously benefited from Google’s decision to prioritize Reddit pages in search and from a data-sharing deal that gave Google access to Reddit’s content. But Google’s AI products can now answer questions using Reddit’s data without sending users to Reddit. Reddit is therefore trying to make users log in, use its app, and receive notifications so that Google cannot be the only owner of the relationship. They compare the situation to Facebook’s long-standing resistance to being fully exposed to search scraping, and to an internal Chrome trade-off in which Google reportedly resisted better bookmark features because they might reduce navigational searches. The larger conclusion is that AI companies can destroy the supply chains that feed them. A Microsoft analysis described the unusual possibility of a successful product killing its own supply chain; the hosts apply that to AI search consuming publishers, Reddit, and eventually YouTube creators. Every platform wants to become one of the few direct destinations users open, rather than merely a discovery layer feeding another service.

The recurring segment “Brendan Carr Is a Dummy” focuses on the FCC chairman’s proposal to auction 25 megahertz of spectrum in the 1675–2020 MHz bands. Carr says the spectrum could support 5G, 6G, direct-to-device satellite services, or hybrid networks, and calls the proposal a first step toward working with federal partners to complete the process “on time, on budget, with no surprises.” The hosts translate the less direct language as a plan to take spectrum currently used by federal incumbents, including NOAA weather systems, and give it to commercial bidders—potentially including satellite companies. They explain that the FCC has moved spectrum before. The 700 MHz auction helped enable 4G after analog television broadcasts were shut down, with the government distributing digital-TV adapters to affected households. But they say the current proposal is especially concerning because the bands support weather detection, hurricane tracking, and weather-balloon data. The phrase “any transition must relocate incumbents” is treated as a polite way of saying that NOAA would have to move or lose access. The hosts connect this to a broader effort by private weather companies to monetize data gathered through public investment. They mention lobbying by AccuWeather and suggest that a fully privatized weather-data pipeline could make basic forecasting more expensive or controlled by commercial interests. Their conclusion is that Carr should explain exactly what the commercial use is and how public weather services would be protected before treating the auction as a routine spectrum proceeding.

The lightning round begins with Anthropic’s S-1 filing, which the hosts call unusually revealing but mostly boring in the way IPO filings usually are. The filing contains large commitments to cloud and infrastructure providers—one estimate discussed on the show is roughly $800 billion over ten years, depending on how the commitments are counted. Anthropic also uses many Google TPUs and Amazon training chips, making it less dependent on Nvidia than some rival frontier labs. The hosts say the company’s financial model requires the broader economy to become deeply AI-enabled and start paying large sums for its products. They question whether every company can sustain a business based on frontier models. OpenAI is also expected to seek public-market funding, Google already has models and distribution, and open-weight models may make capable systems cheaper. The hosts argue that most customers may not need the very best model for ordinary tasks; a sixth-best model could be sufficient. The durable advantage may instead come from products tailored to specific industries. Anthropic’s legal offerings are cited as an example: rather than giving law firms a completely open-ended chatbot, the company is building interfaces and artifacts for particular legal tasks. That approach would require selling into each sector against established enterprise software, which is difficult and slow. Anthropic’s risk factors also contemplate extreme outcomes, including models becoming autonomous or difficult to control. The hosts describe the filing’s overall mood as: the companies may build something world-changing, but they need enormous amounts of money from the economy before anyone knows whether the business works.

The remaining lightning-round items cover Microsoft, robot vacuums, and Anduril. Microsoft is ending Copilot support on smart TVs, including LG televisions. The hosts see the feature as a particularly weak attempt to make Copilot a consumer product: on some TVs it was little more than a shortcut to a web app, and the shutdown notice may have been seen by more people than ever used the feature. They argue that Microsoft’s strongest Copilot position is inside enterprise products such as Excel and Microsoft 365, yet the company repeatedly tries to make it a general consumer assistant. They then discuss Matic’s $1,700 Matic Serene robot vacuum, whose principal new feature is noise reduction. Although both hosts have abandoned robot vacuums after seeing them get stuck or disturb pets and children, they admit that a genuinely quiet robot might solve the most important problem. The device is expensive, but the hosts say Matic’s earlier $1,500 vacuum is widely considered effective because it navigates and understands spaces better than cheaper alternatives. Finally, Nilay describes a dispute with Palmer Luckey over Anduril’s proposed Blue Chip laptop. Early descriptions led some people to believe the laptop would be fully American-made, but when The Verge asked for details, Anduril said the prototype was 32% domestic and that the final percentage was still unsettled. After publication, the company clarified that it expected to reach 65% domestic content at launch—the threshold discussed for the Buy American program. Nilay says The Verge will gladly report that number if Anduril achieves it, but argues that allied-country components do not count as American. He also criticizes the racist abuse directed at him by Luckey’s online supporters, while saying that successfully building a largely American laptop would be a worthwhile achievement.