The Download: the claim shaking up AI and the internet’s agent problem

This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology.

“We were looking at it wrong”: The explosive claim that’s upended the AI world

“I think I have good news about the nature of neural networks.” With those simple words of optimism, Nick Cammarata, a researcher at AI startup Goodfire, kicked off a 26,000-word blog post that has knocked the AI world on its head, sparking wild speculation as well as cautious (and not-so-cautious) celebration.

Cammarata claims that LLMs might be far easier to understand, and thus to monitor and control, than people previously thought. If true, it’s a huge deal. But not everyone is convinced.

I talked to Cammarata the day after his post blew up. We discussed what he thinks researchers are getting wrong, what his work actually shows, and why skeptics aren’t ready to celebrate just yet.

Read the exclusive interview to find out what Cammarata’s claims could mean for AI.

—Will Douglas Heaven

Making the internet better for AI agents might make it worse for everyone else

Until recently, you could split internet users into just two categories: bots and people. But personal AI agents are slowly but surely disrupting that divide by taking on online tasks, from booking flights to reserving hotel rooms. Meta’s Muse alone was downloaded 5 million times in its first month.

The mingling of these agents with human visitors—on an internet designed largely to prioritize people—has gone poorly, with many websites blocking AI agents.

If that demand keeps rising, the internet could be redesigned for agents rather than people. Find out what that could mean for the rest of us.

—James O’Donnell

I’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology.

1 The US Army is trying to get AI agents to hack its enemies
Documents outline a plan to let software find and disable targets. (Intercept)
+ But experts warn they could damage unintended systems. (Gizmodo)
+ AI agents could deepen inequality in developing nations. (Rest of World)
+ “Humans in the loop” in AI warfare is an illusion. (MIT Technology Review)

2 The OpenAI bugging suspect allegedly had Saudi government support
A filing cites royal-family contacts and a military-funded flight. (BBC)
+ The FBI says he had a folder called “Saudi Gov Docs.” (Ars Technica)
+ A judge denied him bail despite a Saudi consulate offer to pay. (CNBC)

3 The US has banned DeepSeek from civilian government systems
Products built using DeepSeek’s models are also banned. (CNBC)
+ The US has sanctioned China’s Y Combinator over AI theft. (Bloomberg $)
+ What’s next for Chinese open-source AI? (MIT Technology Review)

4 Europe is trying to build its own version of Starlink
Four major carriers are exploring a joint bid for spectrum. (Verge)
+ A stratospheric airship has tested a cell tower in the sky. (Ars Technica)
+ Stratospheric internet could take off this year. (MIT Technology Review)

5 Trump officials asked a federal cyber agency to hack voting machines
CISA employees turned down the request. (ProPublica)
+ AI persuasion is entering elections. (MIT Technology Review)

6 US fossil fuel projects are gaining ground again
The White House is backing gas as data center demand rises. (Verge)
+ A climate report row could give polluters a reprieve. (New Scientist $)

7 El Niño is causing sea levels to rise rapidly
Coastal flooding could worsen from Alaska to Chile. (New Scientist $)
+ El Niño could make 2027 even hotter. (MIT Technology Review)

8 Experienced scientists disrupt less, but their ideas last longer
New research shows they produce more lasting breakthroughs. (Nature)
+ A new wave of private labs is challenging US science.(Economist $)

9 Everyone’s buying EVs, except Americans
Drivers don’t want to sacrifice convenience or familiarity. (Atlantic $)

10 Influencer photoshoots are getting the boot at beauty spots
Hotspots are banning them to protect peace and privacy. (Guardian)

Quote of the day

“They all truly hate each other, and they’re all trying to make as much money as they can as fast as they can.”

—JB Pritzker, governor of Illinois, tells The Verge that the fierce competition between AI companies is complicating efforts to regulate the industry.

One more thing

Is fake grass a bad idea? The AstroTurf wars are far from over.

In 2001, Americans installed just over 7 million square meters of synthetic turf. By 2024, that number was 79 million square meters—enough to carpet all of Manhattan and then some. The increase worries folks who study microplastics and environmental pollution.

While the plastic-making industry insists that synthetic fields are safe if properly installed, lots of researchers think that isn’t so. Find out why AstroTurf has ignited heated debates.

—Douglas Main

We can still have nice things

A place for comfort, fun, and distraction to brighten up your day. (Got any ideas? Drop me a line.)

+ Movie nerds, gather round: This interactive database is like IMDB on steroids.
+ FYI for fellow soup fans—pumpkin and coconut milk are a match made in heaven.
+ Now here’s a fun debate topic… What’s the best kind of music to dance to?
+ Bad news, grumpy people! A positive mindset seems to help protect against colds and flu.