L7
Autonomous Hub Infrastructure
Physical access for everyone, everywhere
A standardised, transportable core module — GPU server, graph database, encrypted storage, simulation engine, satellite uplink, solar power, robust tablets — deliverable by truck, ship or helicopter. Three operating modes: Cloud (full connectivity), Hybrid (heavy tasks remote, core local), Island (degraded but functional, no internet required). Three hub types from Basis (dozens of learners) to Regional (thousands plus physical labs).
The module
Containerised data centres are an established market — estimated at roughly $29 billion in 2025. Edge AI has crossed a threshold: quantised 70-billion-parameter models run on a single high-end GPU; models under 9 billion parameters run on smartphones. Satellite internet provides connectivity practically everywhere.
The hub is a standardised, transportable core module: GPU server, graph database, encrypted storage, simulation engine, satellite uplink, solar power and a set of robust tablets. Deliverable by truck, ship or helicopter. The hardware is not the contribution — each component already exists. What L7 contributes is the assembly: the standard module, pre-configured, that lets the rest of the architecture (L1 through L6) run anywhere.
Three operating modes
Three modes define the hub's resilience.
Cloud mode. Full connectivity. Heavy generation tasks run remotely; the hub acts as a thin local cache and an interface point.
Hybrid mode. The cloud handles the most expensive computation; the hub runs core operations locally — profile reads and writes, session state, deterministic simulation engines, the Audit-Agent. This is the expected norm.
Island mode. Degraded but functional. Local models, local graph, local profiles, no synchronisation until connectivity returns. Island mode is the safety net, not the daily mode. But when the internet goes dark — for a day, a week, or longer — learning does not stop.
Three hub types
Three hub types serve different scales. A Basis-Hub serves dozens of learners — a single shipping-container module, suitable for a remote village or a small school. A Standard-Hub serves hundreds and adds a small physical workspace. A Regional-Hub serves thousands and adds physical laboratories and workshops, the places where simulation hands off to apparatus.
The same software stack runs in all three. A learner who began with a Basis-Hub and moves to a city with a Regional-Hub does not migrate profiles or relearn how the system works. The hub scales; the relationship continues.
Bandwidth where infrastructure is weakest
One technical fact deserves precision. The architecture's tier-3 simulation engines — deterministic physics and chemistry — compute the world from lightweight instruction sets. A projectile-motion simulation needs kilobytes of code; a video lecture on the same topic streams hundreds of megabytes.
This advantage applies specifically to deterministic simulations. The Mentor's real-time conversational AI (tier 1) needs cloud connectivity for full capability, which is why hybrid mode is the expected default. But the bandwidth fact stands: the richest form of interactive learning works precisely where infrastructure is weakest. A village with a solar-powered hub and a satellite link can run physics simulations a video platform cannot deliver there.
When the internet goes dark, the hub continues.
Reference
Architecture paper, Section 5, L7. DOI: 10.5281/zenodo.18759134. CC BY 4.0.