One run takes a given spec all the way through definition, an approved Application Definition Package, generated code, a VM provisioned on its own VLAN and VMware host, deployment, and a live health check.
Presets:Proof runCustomer liveQuick smoke
▸ Stack & modelVLAN sandbox pool on ESXi 192.168.8.101 — each run leases its own segment; the infrastructure phase runs first and freezes a checkpoint before any model call.
⏸ Pause for my review after:at each pause: approve, edit the design directly, request changes (the stage re-runs with your notes), or reject to end the run
💡 State every need explicitly — audit trail, alerts, approvals, roles — the design stage maps each stated need to a requirement (and never drops one). Generation follows the Design Guide standards. You will be notified (bell · WhatsApp · webhook) at every gate, failure and go-live — send a test. Full guide: Launching a run yourself.