For the last month, I’ve basically been running two separate versions of Claude: my normal Chat projects (with knowledge files + instructions) and Claude Cowork for the heavy lifting. The problem was Cowork felt like it had no memory of what I was doing—so every switch meant reintroducing myself, moving files around, and wasting time. Anthropic just fixed that. Cowork Projects just dropped, and now you can link Cowork directly to your existing Claude projects: same knowledge files, same instructions, same “brain,” everywhere you work. In this video I walk you through what’s actually changed inside Cowork, how the new “work in a project” flow works, and why I prefer building projects in normal Claude Chat first—then importing them into Cowork. I also build a full “AI consultant” project from scratch: I generate a client intelligence brief using Claude’s deep research, add core knowledge files (business DNA, service playbook, consulting framework), and then write the custom instructions that make the whole system behave like an AI employee. Once it’s imported into Cowork, it can run longer, handle massive file sets locally, remember context inside the project, and output real deliverables—like branded opportunity reports and invoices—without me rewriting prompts every time.