I absolutely love weird apps. Especially the ones built around a single, oddly specific idea. These are the kind of project where you can tell someone made it just because they had a bizarre idea and couldn't let it go until it existed. Up until a few months ago, actually bringing such an idea to reality wasn't something just anyone could do. You needed to commit to learning how to code or find someone who already knew.

By now, you've likely had the realization that AI tools are expensive. You might've realized that by looking at the price of your growing pile of monthly subscriptions. Or the thought might've hit you when you got the polite "you've hit the limit! please come back later" message after using your favorite AI tool for what felt like five minutes.

Anyone who has read the headline is probably wondering, "why would anyone put their Steam library on a USB stick at all?", and I'd think of it as a very fair question. The answer is a little long and winding, but I promise it'll make sense.

Raspberry Pis are fantastic for hobbyist use, but it's really easy to forget that these are powerful SBCs that see scientific use. If you live in an earthquake-prone location and you want to log seismic activity both for scientific purposes and for your own interest, the Raspberry Shake is a full kit that keeps a digital ear to the ground and records any quakes that occur.

This is The Stepback, a weekly newsletter breaking down one essential story from the tech world. For more news about how AI is changing our daily lives, follow Emma Roth. The Stepback arrives in our subscribers' inboxes at 8AM ET. Opt in for The Stepback here.How it startedLong before ChatGPT became a thing, educators and editors frequently used anti-plagiarism tools to see if writers were being honest about their work. These tools work by comparing a written work against a database filled with content from across the web, scholarly articles, and more to check for matching sentences and phrases. Some, like Turnitin, offer a percentage th …Read the full story at The Verge.

As big as Microsoft is, it can make mistakes sometimes. Recently, people noticed that Windows 11 had installed a new OneDrive app on their PCs, even if they didn't ask for it. Worst of all, the app arrived on enterprise PCs, which don't have the right permissions to even open it. Well, it turns out that Microsoft had slipped up and pushed it out too far, too quickly.

Windows 11 has had its fair share of problems over the past few years, and while Windows K2 is doing work to make the operating system more appealing to users, there are still some elements that need some work. For example, Windows Latest claims that the Windows 11 Weather app uses five times more RAM than macOS's Weather app, and the worst part is that Microsoft's version comes with ads bundled in.

It's no secret that XDA has a history with HP's top ultraportable range. We loved it when it was the HP Spectre, and we've loved most of what we've seen of the OmniBook rebranding so far. And this year, HP completely redesigned the OmniBook. The OmniBook Ultra is sleeker, slimmer, and stunning, and I've enjoyed most of my time with it.

I've been writing about AI tools since before ChatGPT became a household name, and something I've noticed since then is that one use case the labs building these tools keep circling back to is travel. It's a trend I've seen across demos, keynotes, press releases, and even the casual examples they drop on social media.

Resizable BAR is one of those features that we've come to appreciate in recent years because it can give a free performance boost in some games. It's essentially a PCIe feature that lets the CPU access the GPU's entire frame buffer at once, rather than in smaller chunks. AMD calls it Smart Access Memory (SAM), but it's the same underlying technology. To set it up, just enable Above 4G Decoding and Resizable BAR in the BIOS settings, and you're good to go. Or so I thought.

Tines launched 3B at the end of July, and its entire pitch is that you can vibe code safely, as credentials never touch generated code. API keys live in connectors and get injected by a proxy outside the execution environment, so the model writing your workflow can't read them even if you ask it to. I said I'd be testing this with a local model, and providing your own model is what the free tier expects after you use your $50 credit allowance. I suspect that most will bring their own cloud-based subscription, but I was especially curious how a local LLM would handle it instead.

Undervolting your GPU is one of the safest and most effective ways to reduce its thermal output, but it is actually the side effect of undervolting that makes it appealing for those in enthusiast circles. You see, silicon has a rather obvious relationship with throughput. If you give it less voltage at the same clock speed, it draws less power for the same throughput, which means it stays cooler and boosts higher for longer.