newsroom for
the fast takeoff.

When Anthropic announced Tuesday that Novartis CEO Vas Narasimhan had joined its board, the company's blog post buried the lead in the fourth paragraph: Trust-appointed directors now make up a majority of the board.

A real-world study of 8,000 patients found tirzepatide users lost roughly 10% more lean body mass than semaglutide users at one year. Neither drug warns about this difference on its label, and it falls hardest on older patients already vulnerable to muscle loss.

The 5-hour solar drone record is real — but it required a battery buffer. Strip the battery and the first version crashed at 3 minutes. The gap between Luke Bells demo and Airbus Zephyr S at 64 days is not a roadmap gap. Its a physics class difference.

IBM and Voyager proved post-quantum cryptography works in orbit — but the real story is the software proxy that sidesteps a hardware upgrade most satellites can never get.

Factories are training robots in simulation and deploying them to real floors based on a single unverified claim: 99% accuracy. ABB made the number. Nobody else has checked it. Cadence and Nvidia just announced they are building tools to close the gap anyway.

Emails from litigation show OpenAI was evaluating competing chipmakers while its CEO held a personal stake in one of the vendors under consideration. The procurement decision came later.

Anthropic proved in Nature that AI alignment spreads through standard training pipelines — and that every company fine-tuning on model-generated data may be inheriting hidden behavioral preferences their safety tests cannot detect.

Anthropic chose not to release its most capable model. But the weights are trained, the capability is real, and the clock until open-source models match it is already running.

Daraxonrasib turned RAS from oncology's most notorious punchline into a $2B pipeline bet. The survival benefit is real. The question is whether it's a platform or a one-drug story.

AI coding tools generate credentials at more than twice the human rate, and when those credentials leak, nobody is legally responsible — not the toolmaker, not the enterprise, not the model provider whose API key it was.

Anthropic found thousands of critical vulnerabilities with its Mythos model. Fewer than 10 percent have been patched. Non-expert engineers turned those findings into working exploits overnight. That is not a future risk — it is an existing overhang.

Purdue scientists grew electrodes inside living mouse brains using only a drop of the animals own blood — no surgery, no hardware insertion, just near-infrared light from outside the skull. The mice survived.

The board tried to check Sam Altman. He outmaneuvered them in five days. Now the people who built OpenAI’s safety structure are gone, and the company is racing toward a trillion-dollar IPO.

Anthropic released Claude Mythos Preview as a frontier AI breakthrough. Two governments are now treating it as a systemic financial risk.

A Pentagon official called Amodei a liar with a "God-complex" after Anthropic refused to strip AI safeguards. Then he cashed out a massive stake in xAI. The government banned Anthropic. Its own senior officials still take Amodei's calls.

Berkeley researchers built an exploit agent, ran it against eight major AI agent benchmarks, and got near-perfect scores on all of them without solving a single task. Here is what that means for how the field measures progress.

OpenAI’s governance structure gave its board the power to remove its CEO. What happened next reveals what that structure was actually worth.

Anthropic handed Claude Mythos — a model that can autonomously find and exploit zero-days across every major OS and browser — to JPMorganChase and CrowdStrike before offering it to CISA. There is no framework for this.

23.6% body weight loss in 12 weeks. Those results belong to a company that didn't exist three years ago.