The network you have.
The network you want.
Close the gap.
agents in the open library
simulators in the constellation
TMF APIs catalogued
cockpit apps
DOIL targets
The Reality Twin
Reality says what the network is. The digital twin says what it should be. The Reality Twin reasons across the two — with context from every operational app — and writes the situation down: "model says all-clear, RF says spoofed, SLA breaches in 18 minutes." That narrative is what the agents act on. Then it goes into the loop.
Read the idea →Deterministic Autonomy
The only way you can actually build this. Put the intelligence where it matters — at build time, acting from experience and authority, reasoning hard while you can still verify it. Make operation the opposite — deterministic, secure, reliable, bounded; no runtime improvisation. A certification loop ties them: automations are synthesised, certified before they run, re-certified as they learn. Autonomy you can sign off on.
Read the idea →The ideas above are built, not slideware. Two surfaces show the machinery whole.
Simulators
10 network twins. Bidirectional sync. TMF v5.0 signal flows. Conversational query. Not a digital twin — a Reality Twin: signals in, models in the middle, agents acting back out.
Enter the constellation →Cockpit
Hangar, Altimeter, Radar, Flight Deck. A cockpit view of the network — composite indexes, compartment models, agent and simulator feeds on one pane of glass.
Step into the cockpit →
DOIL is a language, so it ships like one — $ doilgen compile incident_replayer.doil --target=twin.
Also in the open repo: the agent registry
and a full professional IDE on Monaco
(stable, interconnected, built for production).