It's Agentic QA + auto-provisioning sandboxes. Makes it plug and play to do code reviews that actually run your code instead of looking at it really hard. B/c the agents control all of the environment (ie running all of the services), it's able to collect runtime evidence about pretty much everything.