All updates

Sembix Strata™ · Launch · Build

Background commands: long test runs no longer hold up your team's changes

, Co-Founder & CTO

Share this post:

When Sembi, Strata's AI coding assistant, ran your full test suite or a typecheck, everything else on the job waited until it finished, and on a large codebase that could be sixteen minutes. Your developers waited too, watching a workspace that looked busy but wasn't moving, with no way to see what was running or to stop it early.

Work reaches your team sooner, even when the tests take sixteen minutes

A long test run now moves to the background instead of holding up the work, so Sembi carries on with the rest of a change while the tests run and the change is ready sooner.

Picture a developer who asks Sembi to fix the date validation on a renewal form. Sembi makes the fix and starts the full test suite, which on this codebase takes a quarter of an hour. Before, that quarter of an hour was dead time: Sembi sat on the test run, and the developer either watched or went elsewhere and lost the thread. Now the suite runs in the background while Sembi carries on with the rest of the request.

When the tests finish, Sembi is told and picks up the result, so a failure is dealt with in the same piece of work rather than discovered later. The tests take as long as they always did. What changes is that nothing else waits for them, and the slowest step in a change stops setting the pace for all the others.

The bar under a workspace's message box: two commands running in the background, the full test suite started 8 minutes ago and a typecheck started 4 minutes ago, each with a Stop button

Developers stop waiting out runs that have gone wrong or are no longer needed

Developers can see every long run as it happens and stop it on the spot, so a run that has gone wrong or is no longer needed stops costing time as soon as someone notices.

Say the same developer glances at the workspace and sees two commands running: the test suite, started eight minutes ago, and a typecheck, started four. A bar under the message box lists each one with when it started and its own Stop button. If the developer has just changed direction, and the typecheck is checking code that is about to be rewritten, they stop it there and then. Sembi is told the command was stopped, so it doesn't wait for a result that will never come or report the command as still running.

Before, the only choice was to wait for Sembi to finish everything it had started. The bar doesn't make that call for anyone: a run nobody wants keeps going until a person presses Stop. What it gives them is what they need to decide, which is what is running and for how long.

Getting started

This arrives with your next Strata upgrade, and there is nothing to turn on. Anyone who can see a workspace sees the bar; stopping a run needs permission to manage workspaces in that project. Agree with your developers when a long run should be stopped rather than left to finish, so the people watching know when to step in.

  1. Launch My work: every question an agent is waiting on, in one place
  2. Launch Autopilot Goals: finish a milestone by a date, within a budget you set
  3. Launch Autopilot gets more of your backlog built, and a person still merges every change