Synthetic Minds | Anthropic Has Shipped The AI It Says Should Be Paused.

The Synthetic Minds newsletter offers short daily insights to get you thinking. If you enjoy it, please forward. All signals are powered by Futurwise_. If you need more insights, subscribe to Futurwise andget 25% off for the first three months!_

I have just launched the Intelligence Age Scorecard!It will help you understand how ready your organization is for the Intelligence Age.

Today’s topic: AI & Automation

Mythos Has Shipped. Anthropic Wants A Pause.

Anthropic has shipped Claude Mythos 5. The model has compressed months of engineering into a day and matched skilled human protein designers. The same Anthropic has asked the world to pause frontier AI development.

The same lab shipping the most capable AI ever is asking the world to slow down. The capabilities are real, the concern is real, and the opportunity for organizations is closing fast.

Anthropic has shipped Claude Mythos 5, the most capable model the company has ever made. Stripe used the model to perform a codebase-wide migration on a 50-million-line Ruby codebase in a day. The same work would have taken a team over two months by hand.

In drug design, Mythos 5 matched skilled human protein designers and produced binder candidates for nine of fourteen targets. Anthropic scientists preferred its molecular biology hypotheses over Opus 80% of the time. The model ran novel genomics research over a week of autonomous work.

The Anthropic Institute has published a paper proposing a globally coordinated pause on frontier AI development. More than eight in every ten lines of Anthropic's production code have been written by Claude. Engineer output is running at eight times the 2024 baseline.

That's the news. Here is the signal.

Two things have happened. The lab building the most capable AI has shown us what it can do. The same lab has asked the world to slow down.

The capabilities are not a demo. Stripe compressed months into a day. Mythos 5 matches skilled human protein designers and produces novel scientific hypotheses good enough to advance to experimental evaluation.

Anthropic’s engineers are shipping eight times as much code per quarter as they were a year ago. Or as one engineer put it: "It’s been five months since I last wrote any code myself.”

Now, the lab is asking for a pause because recursive self-improvement, AI building its own successor, has come into view.

While Anthropic is moving a lightning speed, Apple has taken a different approach. Instead of building their own, Apple has rented its AI from a competitor and kept the device, the privacy contract and the user's choice.

The world is converging on a dozen foundation models. Some of those labs will expand into adjacent domains. Anthropic has already put Mythos 5 inside drug-design research at ten times human speed.

Apple has shown one path: rent the model, own the customer, keep the privacy contract. Your company has the same choice to make.

The foundation labs will never own the dentist, the chemist, the architect or the trial lawyer. Those domains are yours.

The lab that built this AI is asking the world to pause. The window for you to combine your domain with abundant intelligence is open, and closing fast.

The question your board should debate is no longer which AI model to buy. It is which domain you own, and how fast you combine it with intelligence its makers want paused.

The Intelligence Age Scorecard

Anthropic has shipped Claude Mythos, its most capable model, and the same Anthropic has asked the world to pause frontier AI development because recursive self-improvement has come into view. WAVE — Watch, Adapt, Verify, Empower — is the question every leadership team owes this pattern: are you still racing to procure a model, or already adapting your domain expertise to intelligence the lab that built it wants the world to slow down?

Take the Intelligence Age Scorecard to benchmark your readiness for the next two quarters, and the next five years.

If this newsletter was forwarded to you, you can sign up here.

Thank you. Mark