Uber president Andrew Macdonald (pictured) says its “hard to draw a line” between AI spending and deliverable features. | Photo: Zed Jameson/Bloomberg via Getty Images After reportedly exhausting its annual AI budget just four months into 2026, Uber is now questioning whether it's actually seeing meaningful returns on its investments. In an interview with Rapid Response, Uber president and chief operating officer Andrew Macdonald said the company isn't seeing a connection between rising token consumption for Claude Code and more useful features being delivered to consumers."That link is not there yet, right? I think maybe implicitly there is more that is getting shipped, but it's very hard to draw a line between one of those stats and, 'Okay, now we're actually producing 25 percent more useful consumer f …Read the full story at The Verge.

This, but no cable. | Photo by Thomas Ricker / The Verge Code in recent Starlink firmware suggests that SpaceX might soon release a version of its smallest internet dish with an integrated battery. A battery-powered Starlink Mini would offer untethered portability for vanlifers, emergency responders, and anyone who wants fast, low-latency internet from almost anywhere on the planet.University researcher Jinwei Zhao spotted a number of strings hinting at the integrated battery in a May firmware release, according to PCMag. The "message DishBatteryStats" line suggests code designed to return specific fields from an integrated battery, including the current state_of_charge. If the dish were simply …Read the full story at The Verge.

The Convention on Certain Conventional Weapons, an international forum that focuses on lethal autonomous systems, is hosted twice a year at the United Nations in Geneva. When Branka Marijan attended in November 2017, she thought the five-day sessions - which dealt largely in hypotheticals, speculating on a world where warfare was fought with killer robots - would be business as usual. After all, this was technology some thought might never be developed, and likely never deployed. That year, she quickly realized, was different. That distant, imagined future was suddenly closer and realer than ever.On the first day, some attendees watched a …Read the full story at The Verge.

Gaming handhelds have absolutely fantastic hardware. Devices like the Asus ROG Ally X are marvels of modern engineering, packing 120Hz variable refresh rate screens, 24GB of RAM, and massive batteries into an ergonomic shell that you can literally hold in your hands and take on the go. However, there's a major visual disconnect: the jarring experience of turning the handheld on. You pass through a beautiful blue animation only to be dropped face-first into a microscopic Windows 11 desktop. You are suddenly using a thumbstick to drag a tiny mouse cursor over a closed window button that is literally three pixels wide and impossible to click on.

Microsoft has been approaching 2026 with a far different mindset than the year before. 2025 was packed full of instances where Microsoft rolled out its AI assistant, Copilot, into as many apps and services as possible. However, after the Windows user base failed to meet Microsoft's enthusiasm for Copilot (going so far as to coin the term 'Microslop' around the turn of the new year), the company has been pulling back its efforts.

Would you listen to magazine articles on Spotify? The streaming platform certainly hopes so, as it's launching a new format for narrated long-form articles, alongside its usual array of music, podcasts, and audiobooks. Starting today, more than 650 articles from publications including Rolling Stone, The Atlantic, Vogue, Variety, Billboard, Vibe, GQ, Wired, Vanity Fair, and Pitchfork will be available in English in regions where Spotify's audiobooks are available.The narrated articles are all under two hours long, and available under the monthly audiobook allowance for Premium users. Spotify says that free users who don't pay for a subscrip …Read the full story at The Verge.

For a while, Windows on Arm devices were in a weird spot. They offered better performance and battery life than typical laptops, but they could only achieve that if the apps supported the Snapdragon architecture. Otherwise, you had to rely on emulation, which didn't really tap into all those juicy benefits of using an Arm chip.

There's a point in AI adoption where the tools stopped being optional. Think about it, how many tools have you tried out at this point or even become reliant on, spanning across domains like productivity, creativity, coding, learning, and so on - a big chunk of this stuff happens in some kind of chat window now. This shift toward AI isn't really news anymore, but with so many new companies popping up all wanting you to try their product, the bill is. And it's not like AI subscriptions are $5 throwaways - some are $20 a pop, if not more, and they all sit alongside whatever else you've already got renewing in the background.

Up until a few months ago, you could've likely gotten away with using free tiers of AI tools exclusively. Most tools didn't have strict limits, and if you hit them, you could just switch to a different tool and keep going. If you hit your Claude limit on the free tier, just switch to your free ChatGPT account. If you ran out there (or weren't getting what you needed with ChatGPT), Gemini was waiting. For a good stretch, that was the workflow for most.

Before a new version of Linux sees public release, it goes through a few rounds of release candidates. These are purely focused on people giving the kernel a spin, spotting bugs, and reporting them so they can be fixed. Linux's founder, Linus Torvald, noticed a strange upsurge in bug reports since version 7.0, and he quickly deduced that it was because people used AI tools to scour the code and report issues automatically.

Home networking can be complicated, especially for the average non-tech-inclined person. As a result, ISPs provide gateway combo units that combine the functions of a router and modem into one box. These boxes typically ship with default settings designed to get the user onboarded and online as quickly and painlessly as possible, and while that's great for them, it results in a network that's running in a less than optimal way. Your network performance, security, and reliability can all be improved by changing just a handful of settings.

The GPU market has been in flux for the better part of 2 years to this point, and buying advice has centered around what's in-stock instead of what's the best deal. When you're trying to ball on a budget, used hardware is always in the conversation. Used flagships can go for what new lower-end models are seen for in storefronts right now, and it's easy for PC enthusiasts to point at the VRAM discrepancy between the two and recommend the older card, but the comparison goes a lot deeper than that. There's certainly trade-offs when it comes to software features, but the flagships of a couple of generations ago have the chops to handle higher resolutions much better than their new low-end counter-parts, and I know which one I'm reaching for every time.