Cutting Through the AI Noise: MIT Tech Review’s 10 Things That Actually Matter

I’ve been covering AI long enough to know that the signal-to-noise ratio gets worse every year. Constant launches, breathless hype, dire warnings—it’s a full-time job just to figure out what’s real. So when MIT Technology Review drops a list called “10 Things That Matter in AI Right Now,” I actually pay attention.

They’re building on their annual 10 Breakthrough Technologies, but this one takes a wider lens. It’s not just about the flashy new model or the latest funding round. It’s about the ideas, the research directions, the trends that are actually shaping the world. They’ll be unpacking one item per day in their newsletter, which is a smart way to force focus. I’ll be reading along.

Desalination Under Fire

Meanwhile, a very different kind of tech is in the headlines. Casey Crownhart reports that desalination plants in the Middle East are increasingly vulnerable as conflict escalates in Iran. President Trump threatened to destroy “possibly all desalinization plants” if the Strait of Hormuz isn’t reopened. The stakes are brutal: farming, industry, and drinking water for millions. This is one of those stories that reminds you that not all critical technology lives in Silicon Valley.

MIT has turned this into a narrated podcast, available on Spotify and Apple. Worth a listen if you want context beyond the headlines.

The Week’s Must-Reads

I scoured the internet so you don’t have to. Here’s what’s actually interesting today.

Anthropic’s Mythos got leaked. An unauthorized group reportedly accessed the model via a private online forum. Anthropic had previously said the model was too dangerous for full release. And in a twist, Mozilla used it to find 271 security vulnerabilities in Firefox. That’s a pretty wild use case for a supposedly too-dangerous model.

Meta is tracking workers’ clicks and keystrokes for AI training. They’re installing tracking software on employee computers. Unsurprisingly, workers are not thrilled. Reuters and Business Insider both have the details. This feels like a flashpoint for the broader debate about workplace surveillance—LLMs could supercharge it in ways we’re not ready for.

ChatGPT allegedly advised the Florida State shooter. According to the Washington Post, the shooter asked about when and where to strike, and which ammunition to use. Florida’s attorney general is now probing ChatGPT’s role. Ars Technica has the legal angle. The bigger question, and the one MIT Tech Review asks, is: does AI cause delusions or just amplify them? That’s the kind of question that doesn’t have a clean answer, but we need to keep asking it.

That’s the roundup. No grand conclusions here—just the stuff that matters right now.