Claude’s new Cowork product is a fine example of Anthropic doing what good product teams do.

They shipped a product (Claude Code), realised that non-engineers found value in it and used that data to inform an entirely experimental new sub-product with Claude Cowork that takes the utility of Code but repackages it into something that is more suitable to its new user segment.

The new product was apparently hacked together in just 10 days and may ultimately fail, but AI is making these types of experiments much easier to run.

This is what Anthropic’s engineer Felix Rieseberg has to say about their product philosophy: