Two pages, two products
Start with the item that matters most to a team, because it changes how people work together. Read-only thread sharing would let a developer hand a colleague a link to one session's transcript without opening the whole project. AlphaSignal describes exactly that, with one restriction: AlphaSignal says threads with local-computer access are excluded from sharing.
Anthropic's documentation describes a product where none of that exists. A project belongs to one user. Its threads cannot be shared with another user. Thread transcripts do not have the share option that other cloud sessions have. There are no organization-level controls during the beta. That is not an omission or a vague paragraph; it is a limitations list, written to tell users what the product will not do.
The daily limit conflicts in a different way: both sources give a number, and the numbers differ. AlphaSignal puts the limit at 200 new threads per day across all projects. The documentation's usage section states an enforced limit of 500 new threads per day across your projects, alongside the plan allowances: threads consume the same plan allowance as any other Claude Code session, they consume it faster because several run at once, and a thread that reaches the five-hour or weekly limit pauses and resumes when the limit resets.
The remaining reported items sit in a third category. The documentation does not describe reminders, the /btw and /fork commands, or a coordinator that posts status updates unasked. That silence does not prove those features are unavailable; documentation can lag a rollout, and features can reach some accounts before the general pages describe them. The sources conflict; their timestamps do not establish which description matches each account's current behavior. What a team can verify today is the documentation, and on the two claims that change how a team works, sharing and the cap, the documentation says the report is wrong.
What is solid enough to use
The underlying system is not in dispute, and it is the part worth a reader's time. The documentation describes the coordinator model in detail. Work arrives in one conversation. Claude decides what becomes a thread, starts each thread with the project's instructions and memory, and tracks every thread it started. Threads run in parallel in the cloud, keep going after the laptop closes, and can be steered from a phone. When a task needs something only the user's machine has, a thread can run on that machine through Remote Control instead, and then it runs only while the machine is awake.
The costs are stated as plainly in the documentation as the capabilities. A new project defaults to Opus everywhere, with high effort for the workers, which draws on the plan allowance fastest. Usage is visible per thread and per model in Project settings, so a team can see which thread is spending its allowance.
Read against that baseline, the reported update is a modest one even if every item in it is accurate. A coordinator that posts its own status updates saves polling. Reminders and branching commands are conveniences. Sharing is the only item that changes who can see the work, and sharing is the item the documentation contradicts. That ordering is the story: the most consequential claim in the report is the one with no primary source behind it.
The practical test is short, and it settles your account rather than the argument. Open a project, open a thread's menu, and look for Share. If sharing appears, verify its audience, restrictions and actual behavior before relying on it; one account's interface does not establish general availability. If it is not there, the documented product is the one you have. Either way, the answer comes from the product and its documentation, not from the number of sites that repeated the report within a day.
Sources
- [1] Anthropic, "Let Claude coordinate ongoing work with Projects" (Claude Code documentation, page updated November 13, 2026, 22:44 UTC)Read source
- [2] Anthropic, "Run agents in parallel" (Claude Code documentation)Read source
- [3] AlphaSignal, "Anthropic Upgrades Claude Code Projects to Run Parallel AI Workers Hands-Free" (published November 13, 2026, 22:51 UTC; reported claims, not confirmed against Anthropic's documentation)Read source