Every release, fix, policy and incident.

Caps can now be set per project as well as per team. Jobs that would cross a cap are held, and the owner is asked.

Training runs now resume from the last checkpoint in under 40 seconds on most pools, down from about three minutes.

A ready-made set of rules that keeps datasets and the jobs that read them inside EU regions.

A new pane shows each region's power, GPUs and queue depth side by side. Press R to open it.

Jobs in one region waited up to 11 minutes to start for 34 minutes. Cause and fix are in the write-up.

The new Terminal, policy files in your own repository and a rebuilt inventory that reads colocation sites.

Audit exports can now be downloaded in either format and filtered by team, region or policy.

Sensible daily caps, alert levels and approvers for teams that are new to shared GPU pools.