AI Digital Twin
A live replica of your workflow, reasoning at the edge.
The Digital Twin Top Model mirrors how your work actually runs, then thinks about it in place — on the device, in the branch office, on the vehicle. No datacenter, no egress, no waiting on a frontier queue.
The twin loop
01
Observe
The twin subscribes to your events through webMCP and keeps a compact state vector.
02
Replicate
Each workflow step becomes a folded nano-prompt, not a service call.
03
Reason
The 1.4k-parameter top model runs locally — 132ms p50, no round trip.
04
Act
Decisions execute back through the bound tool surface, with rollback.
Edge specification
Footprint
38 MB
Cold start
0.4 s
p50 latency
132 ms
Hardware floor
2 cores / 512 MB
Network
Offline-capable
State
Local, encrypted