Start and Monitor an Experiment

Start a checked experiment, retain its run identifier and follow that same run to a final outcome.

Before you start

Confirm the source, workstation, execution mode and preparation requirements. The tutorials use simulation. The 0.2.0 desktop hardware-run action is disabled. Review simulation mode and evidence before starting.

With your AI agent

Start one experiment on the current simulated workstation using the program version you just checked. Report its run identifier and status, then follow its progress without submitting it again.

The agent starts the reviewed snapshot and follows the returned run. To submit first and adjust loading separately, explicitly ask it to submit and show the loading plan instead of automatically starting.

Open this experiment in the desktop and tell me whether it is waiting for preparation, queued, running or finished.

Test example The experiment window shows progress and available controls. This is a native UI test state, not a capture of the tutorial run. View full size.

Understand progress

State or observation	What to do
Waiting for preparation	Complete the remaining loading-plan steps.
Queued or waiting for resources	Inspect the preceding run and resource ownership; avoid duplicate submission.
Running	Follow progress and request pause or cancellation if needed.
Body finished, finalization continuing	Wait for cleanup and resource release.
completed	Review the final report and material records.
Failed, cancelled or needs_reconciliation	Read diagnostics and partial results; do not treat them as a complete success.

An accepted start request is not a completed experiment. If a response is lost, ask the agent to inspect existing runs and the original request. Do not blindly start again under a new request identifier.

Next: Run Controls or Results and History.