Plus, Claude Opus 4.7’s work upgrades.
Anthropic released Opus 4.7 today. While it’s not the company’s most powerful model—that’s still Mythos, which remains in limited release—it’s a stepping stone toward it. Anthropic is using Opus 4.7 to work out the safety guardrails it needs before releasing Mythos more broadly.
So what can Opus 4.7 actually do? Anthropic says it’s meaningfully better at coding—specifically, the hard, multistep work you’d normally need to babysit. It can also apparently see images at three times the resolution of previous models, which matters for anything involving dense screenshots or diagrams. And if you use it for work—slides, docs, interfaces—the outputs should be noticeably cleaner.
Also in today’s newsletter:
- The humidifier you won’t abandon after two weeks.
- The internet has never been this aggressively cheerful.
- Apple’s suing the leaker who spoiled iOS 26 early.
—Whizy Kim and Saira Mueller
THE DOWNLOAD
Is it time to recall Windows 11?
Morning Brew Design, Photos: Adobe Stock
TL;DR: Microsoft released Recall to the public last April—an AI feature that captures your computer screen every few seconds. The release came with a promised security fix after months in previews, but a researcher just proved it’s still exploitable. It’s the latest entry in a long list of user grievances around AI and Windows 11, an OS that’s recently gained market share probably due more to Microsoft killing Windows 10 support than because anyone actually wanted it.
What happened: A cybersecurity researcher has shown that malware can silently extract hoards of data collected by Windows Recall—a supposedly convenient feature that periodically captures what you’re doing on your PC and uses AI to make it all searchable. (This can include anything you’re looking at on your screen, from bank account details to your most regrettable Google searches.) Backlash around its security initially delayed the feature’s launch, and it was released to the public last April with encrypted storage and biometric authentication. Apparently, that wasn’t enough.
The problem: Recall’s data usually sits locked in a secure vault, but malware can trigger Recall to unlock by prompting a real Windows security pop-up where users verify their identity. Once the vault opens, Recall hands your data off to a separate, unprotected process to display it on screen—and that’s when the malware intercepts it.
No fix appears to be coming; Microsoft doesn’t consider this an actual vulnerability, since the OS intentionally allows processes to interact with one another—the same way antivirus software does in order to protect your PC. But such handoffs are typically a momentary exchange with one credential or file, rather than a running record of everything you do on your PC. Some experts argue Recall’s central feature is the real problem and another example of a recurring tension with AI features offering convenience at the cost of security.
Microslop: It’s not just Recall—since its 2021 launch, Windows 11 has faced very public backlash over annoyances like forced Copilot integrations and Start menu ads. The AI features users haven’t asked for—and can’t easily get rid of—even earned the company a new nickname: “Microslop.”
Bottom line: Microsoft’s deployment of yet another unsecured AI feature helps feed the perception many users have of Windows 11—as an OS engineered for Microsoft’s data collection and revenue aims rather than a positive user experience. Some people are eagerly hoping Windows 12 will be better, banking on the folk wisdom that every alternating major Windows release is good. But the Copilot genie is already out of the bottle, and it’ll be very hard to put it back in. —WK
Also at Microsoft…
- The company’s answer to the smash-hit MacBook Neo release: discounts and freebies for students.
The humidifier that doesn’t drive me crazy
Disclosure: Companies may send us products to test, but they never pay for our opinions. Our recommendations are unbiased and unfiltered, and Tech Brew may earn a commission if you buy through our links.
I’ve lived in pretty much the full spectrum of climates and owned a humidifier in most of them. Desert heat, dry apartments with the heat blasting all winter—dry air can irritate your sinuses, wreck your sleep, and do a number on your skin. I’ve used and abandoned more humidifiers than I can count, always for the same reason: Cleaning them was revolting. Which is why I ended up testing the Carepod One Plus, a filter-free, low-tech humidifier with a stainless steel tank designed to be easy to clean.
I tested it in my bedroom for over two months. On my first night in February, a hygrometer—a small device that measures air humidity—showed my room at 26% at 11:42pm before turning on the heater or the humidifier (below the Environmental Protection Agency-recommended 30%–50% humidity range). By 8:50am, with the heat running all night, it was holding at 46%. I ran a similar test with air conditioning a few days ago: My room was at 33% at 10:32pm before I turned it on, and by 6:35am, it was at 46% with the AC still running. I stopped waking up with a dry throat. I no longer lunge for my water glass the second my alarm goes off.
Saira Mueller
Carepod, like most humidifiers, recommends distilled or filtered water, so I boiled tap water every few days, which worked fine but is an easy step to forget on a busy night. Cleaning it took just minutes of my time: The three main parts can go in the dishwasher, get boiled, or be hand-washed. After a few weeks, I noticed a bit of mold on the ring of the oscillator wand—the only place it showed up, and it was easy to clean.
The Good: No filters to buy or replace. The stainless steel design has no hidden corners where mold can hide and grow. Dark mode kills all the display lights for light-sensitive sleepers. Three mist levels and a built-in timer let you dial in exactly how much humidity you want and for how long.
The Bad: At $350, it’s a significant ask for a humidifier with no app, no built-in humidity display, and no auto-adjustment (although not needing filters offsets this cost over time).
Verdict: 
Signal (especially if you’ve rage quit grimy humidifiers before or live somewhere dry enough to feel it in your sleep year-round). —SM
THE ZEITBYTE
The internet is now your “fake-happy” place
Morning Brew Design
If you have a working cringe detector, you probably already knew this: AI writing is pathologically upbeat. According to researchers, by mid-2025, over a third of the text on new websites was generated by AI or AI-assisted. A sentiment analysis found that these sites scored 107% higher on positivity than purely human-written ones, which is totally cool and amazing (and not at all unsettling).
Researchers say LLMs can sound cloyingly optimistic in part because they’re trained to flatter humans—so it’s ironic that this trait tends to be grating to the very crowd AI is optimized to impress. The paper’s own human authors concluded that, thanks to AI, online writing is “increasingly sanitized and artificially cheerful.”
There were some counterintuitive findings, too: Despite all the bot-generated social media comments you might stumble across, there was no statistically significant evidence AI has increased misinformation online. Though, filling the internet with fake chipper people who’ve never had a bad day in their lives is its own kind of gaslighting. —WK
Open Tabs
- Anthropic is planning “a major UK expansion” with a new 800-person office in London, just days after OpenAI announced its “first permanent London office” with space for over 500 employees.
- And crossing the finishing line in third place… Google just released a Mac app for Gemini.
- This one very human (and annoying) part of your job could be even more important given the prevalence of AI.
- Are the footwear companies all right? First Allbirds pivots to AI, and now Reebok is expanding its smart glasses lineup.
- Apple’s backend code reportedly shows four new features coming to iOS 27—including a useful upgrade to Apple Wallet.
- Ukrainian President Volodymyr Zelenskyy says Russian troops surrendered to robots in a very pre-dystopian video post that’s almost 10 minutes long.
- Spotify and three major music labels just won a $322 million judgment against an online database—but they’ll have to find its anonymous creators first.
- Ever wondered what the largest high-res 3D map of the universe looks like? Wonder no longer.
- An Apple leaker dropped details about iOS 26 on YouTube months before its release. Now, Apple’s suing him, claiming a “coordinated scheme to break into an an Apple development phone, steal Apple’s trade secrets, and profit from the theft.” The whole saga is a fascinating read.
Written by Whizy Kim and Saira Mueller