Two execs just bailed on an AI nuclear startup

Fermi, which had been positioning itself as a key power supplier for energy-hungry AI data centers, lost its chief executive and chief financial officer in a sudden, simultaneous departure - with no permanent replacements named and no official reason given.
Why it matters: AI's power problem is real, and nuclear startups were supposed to be part of the fix - losing your top two strategic and financial leaders at once raises serious questions about internal stability at exactly the wrong moment for investor confidence. (source)

A new study found that AI chatbots are providing cancer patients with alternative treatment suggestions - sometimes actively discouraging conventional chemotherapy - and patients are increasingly acting on this guidance without telling their oncologists, with advice ranging from legitimate integrative options to outright misinformation.
Why it matters: People are making life-or-death medical decisions based on outputs from models that carry no medical license, no liability, and zero knowledge of the patient's full history - and regulators are nowhere close to catching up. (source)

Anthropic is developing a model called Mythos, designed specifically to surface and reason about AI safety risks - including uncomfortable scenarios from its own systems - positioning it as a research and evaluation tool rather than a consumer product.
Why it matters: Most AI labs treat safety as a talking point; if Mythos is what Anthropic says it is, it's a direct attempt to institutionalize self-criticism at the model level - though whether that makes it a genuine breakthrough or a sophisticated PR instrument is the question worth tracking. (source)

According to Wired, a growing number of tech executives are training personalized AI agents on their own emails, memos, and recorded decisions - with the goal of simulating their decision-making style, communication patterns, and strategic instincts across the company at scale.
Why it matters: This is either the most efficient leadership innovation in decades or a very expensive way to avoid accountability - and when the AI-CEO makes a bad call, watching who takes the blame will be genuinely instructive. (source)

Cloudflare is redesigning core network infrastructure to support agent-to-agent communication, identity verification, and low-latency routing at scale - a significant architectural shift, given that today's web was built for browsers, not autonomous bots negotiating tasks with each other.
Why it matters: AI agents are only as capable as the infrastructure they run on, and if Cloudflare executes here, it becomes foundational to the agentic AI era - the kind of quiet, unsexy dominance that compounds for decades. (source)

That's a wrap for today. The AI world doesn't sleep, and neither does this newsletter.

Hit reply and tell us which story surprised you most - we actually read every one.

Ready to try every AI model in one place? One subscription gives you access to ChatGPT, Claude, Flux, Sora, ElevenLabs, and 20+ more.

Like what you read?