Unclear It is not clear what decides whether the new wording is used, so it may not apply to every account or model.
Claude Code can run a shell command in the background, so Claude can keep working while it runs. When it does, it tells Claude about it in the tool result, which is the reply Claude gets back after asking to run something. That reply now says Claude will be notified when the command finishes, and should not keep checking the output file or pause to wait.
It also tells Claude to read the output file only:
- when you ask how the command is getting on
- when Claude needs a line the command prints before it can carry on, such as a server's ready line or its port
Before, the reply said Claude would be notified, and told it how to read the file to check on progress. The new wording is used only when the command did not time out, the turn was not stopped, and the reply is not being sent as a message or a final response.
Claude should spend less effort repeatedly checking on background commands and wait for the notice that they are done instead.
It is not clear what decides whether the new wording is used, so it may not apply to every account or model.