Claude Code Projects lands, Cherny stops managing sessions
One conversation splits itself into parallel cloud threads, and two of the people who build it say the session is no longer the unit of work.
The Claude Code account rolled out Projects on desktop and web, in beta for select users. A project is one conversation with Claude that splits the work into threads itself, runs them as parallel cloud sessions, passes context between them, and keeps going when you leave. The Claude account said it is in beta today for select Pro and Max users in cloud sessions, coming to all Claude users soon.
The posts worth reading are about the loop it replaces, not the feature list. Boris Cherny said Projects changed not only how he interacts with Claude but how he codes, and that it is where he does a ton of his coding now. Asked about it in a reply, he added that Claude does it for him and there is no need to tell Claude to do it anymore.
Thariq described the shape underneath. One agent per project manages memory and spins off subagents for tasks, and you can ask it to be proactive or to work on a schedule. His verdict was that it feels a lot nicer than a bunch of sessions.
The same direction from the local side
Thorsten Ball is pushing at the same problem with different plumbing. He posted that it is time to leave one `amp --no-tui` running on a Mac mini and forget about it, because it now serves every repo on the machine and keeps itself updated. No orchestrating agent in that picture, just a daemon that is always there.
Time to leave one `amp --no-tui` running on your Mac mini and forget about it.
For working engineers, the common bet is that you stop being the scheduler. If a coordinator agent owns memory and fan-out, the unit of work is no longer the terminal session you keep open and feed, it is the project you check on. That also moves where things go wrong. A session you babysit fails in front of you, while threads that keep working after you close the laptop fail somewhere you are not looking.
Worth keeping the certainty straight. These are claims from the people who build the tools, on their own workflows. Projects is in beta for select users, so nobody outside that group has put it on real work yet, and there are no third party results to compare against the long-lived single session that most people are still running.
The older habit these replace, tracking what an agent changed closely enough to defend it, is a separate question. Gary Bernhardt's answer was to cut most agent diffs to a quarter of their size.
I stopped managing sessions. I just send thoughts as they come, Claude splits them into threads, and the project remembers how I work.
It has one agent per project that manages memory and spins off subagents for tasks.