1 · Pick 2 · State 3 · Ship Change behavior later
1 · Pick the hardware
biRT runs on a runtime host you provide — typically a small GCP VM, scaled to your site. The host carries the intent; the operator-facing devices already in the building connect over the local mesh. No new operator-side hardware to buy.
2 · State your intent (operator to NOUS)
“I run a 60-seat dine-in. Take my menu, route orders from the front to the kitchen, settle checks at the table. We're open 11 to 10.”
3 · Ship (NOUS confirms)
“Manifest signed and broadcast over the local mesh to the phone-KDS and POS tablets. Applied at next tick. Dine-in is live — speak any name change for any item or hour any time.”
4 · Change behavior later (operator to NOUS)
“Hey NOUS, we're closing the second grill station for maintenance — should be back up in a few days. Push grill-2 orders over to grill-1, prioritize salads and bowls on the kitchen displays, and have the front-of-house gently steer guests away from grill-heavy items until I tell you we're back.”
5 · What happens
Signed manifest goes out over the mesh to every runtime in the building — KDS, POS terminals, FOH tablets. Applied at the next sync tick. No code shipped. No restart. No outage. Reverted just as fast when the operator says it's back.
1 · Pick the hardware
One GCP ELITE-tier compute instance (c2-standard-8) for the plant-wide controller — bulk translation, manifest broadcast, alert routing all on the metal. Each line station's existing operator hardware connects over the mesh.
2 · State your intent (operator to NOUS)
“Five lines, plant A. Track throughput. Alert me when any line drops below target by 10%. If a line slows, route surplus to lines with headroom. Shift starts 6am.”
3 · Ship (NOUS confirms)
“Manifest signed and broadcast from the controller to every line phone. Applied at next tick. Throughput board is live — speak any new threshold any time and I redeploy.”
4 · Change behavior later (operator to NOUS)
“NOUS, line 3 is running half speed today — QA is sampling. Don't trip the under-throughput alert, and route the surplus to lines 4 and 5 if they have headroom. End of shift, go back to normal.”
5 · What happens
Manifest signed and broadcast to all five line runtimes. Applied at next tick. Alert thresholds shift; routing kicks in only if 4 and 5 have capacity. The runtimes themselves do not change. Auto-reverts at shift end.
1 · Pick the hardware
biRT runs on a runtime host in dispatch — a GCP VM or a local x86 controller, your call. Field crews carry whatever ARM hardware your team already standardizes on as the mesh terminal: biRT-A on the device when the bootloader is locked, biRT bare metal when it isn't. They mesh over BLE, LoRa, WiFi, and serial.
2 · State your intent (operator to NOUS)
“Dispatch jobs by zip. Page the storm-response team on weather alerts. Suppress non-emergency tickets in active storm zones until the all-clear.”
3 · Ship (NOUS confirms)
“Manifest signed and broadcast across the mesh. Applied at next tick. Dispatchers see the new routing now. Crews stay in sync without cloud uptime.”
4 · Change behavior later (operator to NOUS)
“NOUS, the storm is hitting the southern region. Postpone all non-emergency dispatches in zip 30500–33999 for the next 12 hours. If anyone in that area calls about an active outage, page the storm-response team instead of the regular crew.”
5 · What happens
Manifest pushes to dispatch and customer-portal runtimes across the affected region. Applied immediately. Auto-reverts in 12 hours unless extended. No code changes anywhere.
Honest note

These three flows are illustrative of the design, not transcripts from a production system today. The runtime, the manifest pipeline, and the mesh transport are shipped. The conversational manifest authoring layer is in active development. We will update these flows as each layer hardens, and we will publish real numbers the first time a customer ships.