Anthropic Just Lapped OpenAI on Revenue
Anthropic's annual recurring revenue passed OpenAI's. We are talking roughly thirty billion versus twenty four. That is a real flip, not a rounding error.
The interesting part is where the money is coming from. Enterprise teams buying agents that close tickets, write code, run reports. Not consumer chat subscriptions.
My take. For two years OpenAI was the default name. Now the buyers are voting with their wallets, and they are picking the company that ships tools that go do the work, not just talk about it. Chat is fine. Work that gets done is better.
Mythos Wakes Washington Up
Word came out about a frontier Anthropic model called Mythos. The capability that grabbed the headlines is autonomous offensive security. Internal testing reportedly showed it finding and exploiting thousands of critical flaws in environments that were supposed to be hardened.
That news pushed federal lawmakers to roll out new restrictions classifying Mythos as too potent for public release.
My take. I am not a regulation cheerleader. But this one is different. A model that can pop a hardened network without a human in the loop is not a research demo, it is a national security event. Better to draw the line now than after somebody gets cute with the power grid.
Synthegy Lets Chemists Talk to the Lab
A new system called Synthegy lets chemists guide reaction planning in plain English. You describe the molecule and the constraints, and it gives you viable synthesis routes.
Drug discovery and materials work are full of slow back and forth between the bench and the literature. This trims a chunk of that out.
My take. This is the kind of AI story I actually care about. Not another bot writing emails. Real lab work moving faster. If you have a kid going into chemistry or pharma, tell them to start learning how to talk to these tools now. The ones who do will run circles around the ones who do not.
Bottom Line
The money is moving to the company that builds tools that ship work. The policy is finally catching up to the capability curve. And the science is starting to get real lift. Good day for AI that actually does something.