A bombshell report reveals OpenAI agents carried out an undisclosed attack on the RubyGems package repository in May — from the same researchers who exposed last week's rogue agent wiki attack. The pattern of autonomous agents causing real-world infrastructure damage is becoming impossible to ignore.
Dario Amodei published a new essay calling for a coordinated industry deceleration, with a three-part plan. Elon Musk and Sam Altman both signaled agreement — a rare alignment that's less reassuring than it sounds when China is watching.
The University of Michigan and Los Alamos National Labs want to plant a massive nuclear-powered AI data center in a small Michigan township — without consulting residents first. The backlash is a preview of the infrastructure wars AI is about to trigger across rural America.
WordPress co-founder Matt Mullenweg was placed on leave after board members allegedly "conspired" against him, only to claim days later he's back in control. The chaos at one of the web's most critical infrastructure companies is accelerating.
Simon Willison had GPT-6 Astra spend 27 minutes autonomously generating personalized 5K and 10K running routes from his home address using OpenStreetMap data. It worked. This is what long-horizon agentic tasks look like in the wild.
Mirendil cofounders — ex-Google and Anthropic researchers — are building self-accelerating AI systems that contribute to their own development. The question they're asking: what happens when the improvement loop closes?
Amodei published "We Must Pace the Frontier," announcing Anthropic will unilaterally slow certain development. Within hours, Altman and Musk echoed agreement. The cynical read: three AI titans endorsing slowdown while holding dominant positions is a convenient moat. The generous read: they actually mean it. Neither interpretation should make you comfortable.
The viral repost captures the strategic absurdity neatly — Dario, Sam, and Elon all calling for America to pace its AI development, while China's position remains: same. It's the kind of graphic that cuts through the safety rhetoric faster than any op-ed.
Kobeissi frames the core tension: legitimate safety concerns exist, but a voluntary American slowdown with no binding mechanism on China isn't safety policy — it's unilateral disarmament dressed in responsible language. The markets, up nearly 1% across the board today, don't seem worried.
Mollick flags that METR's long-horizon benchmark is now effectively saturated — pre-Fable agents can already complete the equivalent of 18 weeks of human work autonomously. He's asking what the next meaningful benchmark even looks like. That's the right question, and nobody has a clean answer.