Claude Cowork now remembers your regular chats: memory becomes the key currency of AI assistants

Anthropic updated Claude Cowork: the assistant now remembers the content of a user's regular chats and draws on them in new work. The practical meaning — the eternal "explain everything again" disappears: the helper knows what you worked on, what decisions were made and how you like to write.

The vendor race around memory shows where an AI assistant's value really lives: not in the model, which is roughly equally clever for everyone, but in accumulated context. The more a helper knows about your tasks, the harder it is to leave — memory becomes both convenience and lock-in.

For business this raises a question worth asking before mass adoption: whose context is it? If working memory accumulates in employees' personal accounts, then a departing manager takes along the assistant "trained" on their tasks, and corporate details settle in someone else's cloud. The right configuration is when context belongs to the company: knowledge about processes and clients lives in the corporate space, access is managed, and a staff change does not zero out what was accumulated. AI memory is an asset; watch whose balance sheet it lands on.