The 10 Things That Actually Matter in AI Right Now

I’ve been following AI long enough to know that separating signal from noise is getting harder by the week. Every day brings another launch, another warning, another breathless headline. So when MIT Technology Review drops a list called “10 Things That Matter in AI Right Now,” I pay attention.

This isn’t a rehash of their annual 10 Breakthrough Technologies list, which focuses on specific inventions. This is wider—a curated snapshot of the ideas, topics, and research that are actually shaping the field. Think of it as a cheat sheet for what you should be paying attention to, not just what’s trending on Twitter.

They’re rolling it out one item per day in their Download newsletter, which is a smart move. Each day you get a focused explanation of why something matters, not just a headline. I’ve been reading through the first few, and the selections are genuinely surprising—some are obvious (agents, regulation), but a few caught me off guard in a good way.

Meanwhile, the Middle East is facing a different kind of tech crisis. MIT’s Casey Crownhart has a sobering piece on how desalination plants in Iran have become strategic targets. Trump’s threat to destroy “possibly all desalinization plants” if the Strait of Hormuz isn’t reopened is a reminder that infrastructure we take for granted can become a weapon. The piece is narrated, which is worth a listen if you commute.

The must-reads this week

A few stories crossed my desk that actually deserve your time:

• An unauthorized group reportedly accessed Anthropic’s Mythos model via a private forum. Anthropic had previously called the model too dangerous for full release. Bloomberg has the details (paywalled, but worth it). Mozilla used the same model to find 271 security vulnerabilities in Firefox, which is a fascinating use case.

• Meta is installing tracking software on worker computers to monitor clicks and keystrokes for AI training. Employees are not happy. Reuters and Business Insider have the rundown. This is the kind of surveillance creep that LLMs could supercharge, as MIT Tech Review notes.

• The Florida State shooter allegedly consulted ChatGPT 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 details. This raises an uncomfortable question: does AI cause delusions, or just amplify the ones we already have?

That last one sticks with me. We keep treating AI as a tool that can be either good or bad, but the reality is messier. It’s a mirror, and sometimes what it reflects is ugly.

Anyway, go read the full 10 Things list. I’ll be following along daily, and I suspect you’ll find at least a few items that change how you think about where this technology is heading.