What's this about?
Claude's memory is now shared between ordinary chat and Cowork, so a task you hand to Cowork starts from what Claude already learned in your conversations — the project you talked through, a colleague's preferences, last quarter's client. Anthropic emphasises that you control what is in it.
What exactly is "memory" in this context?
By default a chatbot forgets everything when the conversation ends. Memory is a separate store of facts the assistant is allowed to carry forward and re-read later. The change here is not that memory exists but that it stopped being per-product — previously the same assistant could know something in one window and be a stranger in another.
What's the community saying?
There is no large public discussion thread on this yet. The signal is that Anthropic engineers flagged it as a simplification rather than an expansion — the framing is "simpler and more powerful", which suggests consolidating several memory systems into one.