Nuclear Waste and AI Agents: Two Problems We Can’t Ignore

The Nuclear Waste Elephant in the Room

Nuclear energy is having a moment. Public approval is up, Big Tech is pouring money into it, and for once, it has bipartisan support. But there’s a problem we keep kicking down the road: what do we do with the waste?

Every year, US reactors produce about 2,000 metric tons of high-level waste. And right now, there’s nowhere permanent to put it. That’s not a sustainable situation, and the urgency is growing.

Casey Crownhart over at MIT Technology Review’s The Spark newsletter laid this out clearly. The irony is hard to miss: we’re finally getting serious about nuclear power, but we’ve been ignoring the waste problem for decades. Yucca Mountain was supposed to be the answer, but that project stalled out years ago. Now we’re scrambling.

This isn’t just a technical challenge—it’s a political and regulatory one too. Finding a site, getting community buy-in, and navigating the legal hurdles takes longer than building the reactors themselves. If we’re serious about nuclear as a climate solution, we need a serious waste plan. Period.

Orchestrated Agents: The Assembly Line for White-Collar Work

On the AI front, Will Douglas Heaven makes a point that’s been rattling around my brain: when people say AI will transform industries, what they really mean is AI agents. ChatGPT proved AI can talk. But to actually change the world, it needs to do stuff.

The real shift happens when agents work together. Tools like Codex and Claude Cowork are early glimpses of multi-agent systems coordinating to handle complex tasks. Heaven compares this to what assembly lines did for manufacturing—and honestly, that’s not an exaggeration.

But here’s the thing I keep coming back to: with that power comes real risk. When agents start making decisions in real-world systems, the margin for error shrinks. A bug in a chatbot is annoying. A bug in an agent managing supply chains or financial transactions is a disaster waiting to happen.

Agent orchestration is one of MIT Technology Review’s 10 Things That Matter in AI Right Now, and for good reason. This isn’t sci-fi anymore. It’s happening, and we need to think hard about how we deploy it.

Mirror Life: A Cautionary Tale

There’s a wild story from Stephen Ornes about “mirror” bacteria that’s worth your time. Back in 2019, scientists proposed building synthetic microbes with mirror-image proteins and sugars. The idea was to unlock new insights into cellular biology and drug design.

Now, many of those same scientists have reversed course. They’re worried that mirror organisms could trigger a catastrophic event threatening all life on Earth. That’s not hyperbole—it’s a genuine concern from people who know this stuff better than anyone.

The MIT Technology Review Narrated podcast covered this, and it’s the kind of story that keeps you up at night. Not because it’s alarmist, but because it shows how quickly scientific enthusiasm can turn into existential caution.

What I’m Watching

Elon Musk testified yesterday in the OpenAI trial, claiming Sam Altman “stole a charity.” That’s going to be a fascinating legal battle to follow. The implications for AI governance and open-source versus closed-source models are huge.

But for now, the big takeaways are clear: nuclear waste needs a real solution, AI agents are coming whether we’re ready or not, and sometimes the most exciting science is also the most dangerous.