From Cellular to Civic

A. C. Applebee and A. D. Masters

Abstract

The telecommunications industry spent two decades solving a coordination problem that civic technology still pretends is unsolvable: how to run a network of millions of independent nodes, across heterogeneous hardware and hostile conditions, without a single controller that can be bought, compromised, or shut off. The answer is the Open Radio Access Network (O-RAN) architecture, ratified by a 30-operator alliance and documented across thousands of pages of public standards (O-RAN Alliance, 2023; Habibi et al., 2023). (Applebee & Combe, 2026, "From Cellular to Neural") applied that architecture to sovereign AI agents and built the working system. (Applebee & Combe, 2026, "The Invisible Network") built the Bluetooth Low Energy mesh that carries OMXUS traffic without servers, towers, or ISPs. This paper closes the loop. It shows that OMXUS — its identity layer, its voting protocol, its emergency response, its quarterly headcount — is a direct architectural descendant of O-RAN, and that the coordination problems OMXUS solves have already been solved at the scale of global cellular networks. Humans are the base stations. Votes are the traffic. The mesh is the radio access network. The SMO is the referendum calendar. The RIC is the fraud-detection layer. None of this is metaphor. It is the same architecture, applied to a different physical layer, serving a different population of nodes. The claim that "civic infrastructure requires a central authority" is exactly as true as the claim that "cellular networks require a central controller." It was false in 2018. It is false now.

Keywords: O-RAN, SMO, RIC, civic infrastructure, BLE mesh, decentralized coordination, direct democracy, sovereign identity, emergency response, OMXUS

1. Three Papers, One Argument

OMXUS exists as four pillars — identity, voting, safety, and a sovereign communication network — and the research series has described each of them in isolation. (Applebee & Combe, 2026, "Be In The Same Room") explained the vouch ceremony. (Applebee & Combe, 2026, "Swiss Direct Democracy") explained quarterly referendums. (Applebee & Combe, 2026, "Wanted Attention for Unwanted Results") explained the three-second emergency hold. (Applebee & Combe, 2026, "The Invisible Network") explained the BLE mesh that carries all of it.

What the series has not done, until now, is explain the coordination layer that ties them together. The Non-Real-Time RAN Intelligent Controller. The Service Management and Orchestration Framework. The rApps and xApps. The A1, O1, O2, E2 interfaces. These are not terms borrowed for aesthetic reasons. They are the names of the components that make the whole thing function — borrowed, exactly and deliberately, from the cellular industry's open standard for network coordination.

(Applebee & Combe, 2026, "From Cellular to Neural") made this argument for AI agents. It showed, with 7,980 lines of production Go code, that the O-RAN pattern applied to autonomous agent orchestration works at 10,000+ events per second, on commodity hardware, without cloud. That paper closed one escape route: "AI agents cannot coordinate without centralized cloud infrastructure."

This paper closes another: "Civic infrastructure cannot coordinate without government infrastructure." Same architecture. Different traffic. Same conclusion.

2. What O-RAN Actually Solved

A modern 5G network manages over 100,000 base stations per operator, millions of simultaneous device connections, and thousands of handoffs per second across multiple spectrum bands and quality-of-service classes (O-RAN Alliance, 2023). No single controller can run this. The monolithic architecture failed by the late 2010s — too slow, too expensive, too locked to single vendors.

The O-RAN Alliance, founded in 2018 by a consortium of operators including AT&T, China Mobile, Deutsche Telekom, and NTT DOCOMO, published a disaggregated architecture that split the problem into layers (Habibi et al., 2023):

The Service Management and Orchestration Framework (SMO). The slow, deliberate layer. It holds the inventory of every network function, defines policy, manages the lifecycle of components, and runs the non-real-time control loops. Decisions here operate on timescales of seconds or minutes.

The Non-Real-Time RIC. Embedded in the SMO. Runs modular applications called rApps that provide policy recommendations, data analytics, and AI/ML inference. Feeds decisions down to the Near-RT RIC via the A1 interface.

The Near-Real-Time RIC. Runs at the edge of the network. Hosts xApps that execute control loops in 10 milliseconds to 1 second. Manages fine-grained optimization — traffic shaping, handoffs, interference mitigation.

The radio access layer itself. The O-CU (central unit), O-DU (distributed unit), and O-RU (radio unit). These are the things that touch the air. They report metrics upward and receive policy downward.

The critical move was interface standardization. A1 between the RICs. O1 between the SMO and everything else. O2 to cloud infrastructure. E2 between the Near-RT RIC and the radio hardware. Once the interfaces were open and standardized, no single vendor could own the stack. Competition arrived. Interoperability arrived. Lock-in ended.

The second critical move was that nothing phones home. Each O-Cloud site, each cellular site, each radio unit runs autonomously and coordinates with its neighbours and its RIC. There is no single server in Menlo Park that every base station in Australia dials back to for permission.

This is the architecture that runs every major 5G network on Earth.

3. The Mapping

OMXUS borrows the pattern directly. The physical layer changes, the traffic changes, the population of nodes changes — but the architectural problem is the same, and so is the solution.

| O-RAN component | OMXUS equivalent | What it carries |
|---|---|---|
| O-RU (radio unit) | A verified human with a smartphone | Presence, signatures, votes |
| Cellular site | A cluster of humans within BLE range (≤400m) | Peer-to-peer relay |
| O-DU (distributed unit) | The BLE mesh router running on each phone | 7-hop TTL, deduplication cache |
| O-CU (central unit) | The device's OMXUS app — identity keypair, signing, local policy | EIP-712 signed messages |
| Near-RT RIC | The on-device safety and fraud layer | SOS propagation, vouch validation, proximity attestation |
| xApps | Emergency response, proximity verification, conflict mitigation | Real-time decisions on the 10ms–1s scale |
| Non-RT RIC | The referendum and policy layer | Quarterly votes, influence budget rollover, vouch chain analysis |
| rApps | Vote, Vouch, SOS, ViewSwap, headcount reconciliation | Slow, deliberate decisions on the second-or-longer scale |
| SMO Framework | The open OMXUS protocol itself | The calendar, the ruleset, the public ledger |
| A1 interface | How policy flows from the referendum layer to the device | "These are the active referendums. This is your influence budget." |
| O1 interface | How device state flows up to the protocol layer | "This is who I vouched for. This is my vote. This is my presence proof." |
| O2 interface | How the protocol reaches external services (verification API, revenue split) | The fractional-cent verification check |
| E2 interface | How the on-device safety layer reacts to mesh events | SOS received → alert propagates → nearest verified human responds |

Read the table twice. Every component on the left is real, standardized, running in production somewhere in Australia on a Telstra tower right now. Every component on the right is a direct analogue that OMXUS already implements or is implementing. The architecture is not being invented. It is being inherited.

What differs is the physical layer. O-RAN operates on licensed spectrum through amplified radio equipment mounted on towers. OMXUS operates on the unlicensed 2.4 GHz band through the Bluetooth Low Energy radios already present in 6.8 billion smartphones (Applebee & Combe, 2026, "The Invisible Network"). A tower costs between $200,000 and $400,000 to deploy. A human walking down the street costs nothing.

4. Why This Framing Matters

The objection to OMXUS has always arrived in the same shape. "You cannot run an identity system without a central authority." "You cannot run an election without a central tabulator." "You cannot coordinate emergency response without a dispatch centre." "You cannot verify humans without a government database."

Every one of these objections assumes a specific architecture — monolithic, centralized, controlled by a single operator — and then observes, correctly, that such an architecture cannot be run without a central authority. The objection is internally consistent. It is also aimed at a system that OMXUS is not.

O-RAN is the refutation. Cellular networks are more complex than civic coordination. They manage more nodes, more traffic, more interference, more edge cases, more adversarial conditions. They do it across heterogeneous hardware from dozens of vendors under hostile electromagnetic conditions, and they do it in real time at millisecond latencies. And they do it without a single controller. The entire industry abandoned that architectural assumption a decade ago. It was cheaper, more reliable, more competitive, and more secure to run coordination as a layered, interface-standardized, peer-federated system.

If the cellular industry could disaggregate a problem this hard, the claim that civic infrastructure is too delicate to disaggregate is not a technical claim. It is a political one.

The reason civic infrastructure has not been disaggregated is not that the engineering is hard. The engineering has been done. It is published. It is open. It is running in every country on Earth. The reason civic infrastructure has not been disaggregated is that the parties currently running it — governments, platforms, certifying authorities, private prison companies, pharmaceutical lobbyists — do not want it to be. The centralized architecture is not a technical requirement. It is a revenue model.

5. What OMXUS Inherits

Because OMXUS borrows the O-RAN pattern rather than inventing a new one, it inherits two decades of industry work for free.

Standards inheritance. The interface definitions — A1, O1, O2, E2 — are public, documented, and have reference implementations. The SMO Framework specification (Habibi et al., 2023) runs to dozens of pages of rigorously defined message formats and state machines. OMXUS does not need to design its coordination protocol from scratch. It needs to map it. The mapping in Section 3 is the start of that work.

Tooling inheritance. Every vendor that builds for O-RAN — Nokia, Ericsson, Samsung, Mavenir, dozens of startups — produces open-source and commercial tooling for the interfaces. Test harnesses, conformance checkers, policy validators, rApp development kits. Most of this is directly reusable by anyone writing a civic application against the same pattern.

Pattern inheritance. The hardest problems in coordination — how to route around a failed node, how to reconcile conflicting policies, how to age out stale state, how to detect and quarantine compromised nodes — have been solved inside O-RAN and published. OMXUS does not need to rediscover these patterns. It needs to apply them, with adjustments for the civic population of nodes (humans are slower than base stations; humans can also refuse consent, which base stations cannot).

Code inheritance. (Applebee & Combe, 2026, "From Cellular to Neural") already produced 7,980 lines of production Go implementing SMO, Near-RT RIC, Non-RT RIC, and rApp frameworks for the AI agent case. The civic case reuses most of this. The Go binary, the three-node mesh, and the chat.omxus.com relay that currently form the OMXUS prototype are running variants of the same components. This is not speculative architecture. It is operational code.

6. The Two Monkeys Already Have the Architecture

The experiment is famous. Two monkeys do the same task. One gets a grape. The other gets a cucumber. The cucumber monkey, on seeing the grape, throws the cucumber back. Any organism with a sense of fairness responds the same way. The unfairness is not complicated. The system giving out cucumbers is.

For most of the twentieth century, the system giving out cucumbers was defensible as a technical constraint. Someone had to run the database. Someone had to count the votes. Someone had to dispatch the ambulances. Someone had to verify the identities. The centralized architecture was not preferred — it was the only one that worked.

That ended in 2018 when the O-RAN Alliance published its first specifications. It ended again in 2023 when those specifications were extended to cover interoperability across 3GPP, ETSI, and O-RAN components (Habibi et al., 2023). It ended for AI agents in the working implementation of (Applebee & Combe, 2026, "From Cellular to Neural"). And it ends, for civic infrastructure, the moment OMXUS ships.

The architecture is already built. The standards are already written. The code is already running. The only remaining question is whether the civic population — 26 million Australians, 6.8 billion humans worldwide — chooses to run on the same pattern that the 5G network in their pocket already runs on.

The cellular industry solved this. The AI agent industry is solving this. The civic use case is the easiest of the three. The nodes are slower. The traffic is lighter. The latency requirements are looser. Four referendums a year is not ten thousand handoffs a second.

The two monkeys are still in their cages. The architecture to open the cages already exists. Someone just has to turn it on.

References

Applebee, A. C. & Combe, L. N. (2026). Be In The Same Room. OMXUS Press Series.

Applebee, A. C. & Combe, L. N. (2026). From Cellular to Neural. OMXUS Press Series.

Applebee, A. C. & Combe, L. N. (2026). The Invisible Network. OMXUS Press Series.

Applebee, A. C. & Combe, L. N. (2026). The Switchboard. OMXUS Press Series.

Applebee, A. C. & Combe, L. N. (2026). Swiss Direct Democracy. OMXUS Press Series.

Applebee, A. C. & Combe, L. N. (2026). Wanted Attention for Unwanted Results. OMXUS Press Series.

Habibi, M. A., Yilma, G. M., Costa-Pérez, X. & Schotten, H. D. (2023). Unifying 3GPP, ETSI, and O-RAN SMO Interfaces: Enabling Slice Subnets Interoperability. IEEE Future Networks World Forum, Baltimore, MD, November 13–15, 2023.

O-RAN Alliance. (2023). O-RAN Architecture Description v9.0. O-RAN.WG1.OAD-R003-v09.00.

Published by OMXUS Press. Creative Commons Attribution 4.0 International (CC BY 4.0).