Published ahead of Forbes Under 30 Summit, April 19–22, Phoenix.
You are about to walk into a room full of the most ambitious founders on earth.
You will hear a lot of pitches about AI. Most of them will describe systems that work beautifully at small scale and silently choke as they grow. Not because the founders are wrong about the problem. Because every one of them is building on the same architectural ceiling — and almost no one in that room has named it yet.
This article names it. And it points to a discovery that changes the math.
The Problem Nobody Is Pitching
Here is the architecture assumption buried inside nearly every AI company presenting at Forbes Under 30 this year:
Intelligence lives in a central place.
A model. A server. An orchestrator. A data lake. Something that ingests everything, processes everything, and sends answers back out.
This assumption is so pervasive it has become invisible. It is in LangChain. It is in AutoGen. It is in every RAG pipeline, every federated learning proposal, every "intelligent platform" pitch deck.
And it has a hard ceiling.
When you centralize intelligence, latency grows linearly with the number of agents. Compute grows linearly with the number of data sources. And the orchestrator — the thing that makes it all work — becomes the thing that makes it all fail. One bottleneck. One point of failure. One constraint that never goes away no matter how much you scale.
This is not a compute problem. This is an architecture problem.
You can throw GPUs at it indefinitely. The ceiling stays.
The Math That Changes Everything
In 2025, Christopher Thomas Trevethan discovered something that had never been formally proven before: a complete distributed architecture in which intelligence scales quadratically while compute scales logarithmically.
This is not a marginal improvement. This is a phase change.
Here is the math:
- N nodes in a network = N(N-1)/2 unique synthesis opportunities
- That is Theta(N^2) — quadratic growth
- Each node pays only O(log N) routing cost — logarithmic growth
- 10 nodes => 45 synthesis pairs
- 100 nodes => 4,950 synthesis pairs
- 1,000,000 nodes => ~500 billion synthesis pairs
The intelligence compounds as the network grows. The compute per node barely moves.
Trevethan's discovery is called the Quadratic Intelligence Swarm (QIS) protocol, covered under 39 provisional patents. It is not an AI model. It is not a framework. It is a discovery about how intelligence naturally wants to flow when you stop trying to centralize it.
What the Architecture Actually Does
The key insight: instead of routing raw data to a central processor, QIS routes pre-distilled outcome packets to semantically similar nodes.
The complete loop:
- Raw signal arrives at an edge node (a hospital, a sensor, an autonomous vehicle, a clinic)
- Local processing runs on-device — raw data never leaves
- The result is distilled into an outcome packet (~512 bytes) — what worked, what failed, what was observed
- That packet gets a semantic fingerprint — a vector representation of the problem it solved
- The packet is routed to semantically similar nodes — other edges facing the same problem — via any efficient mechanism (vector database, DHT, pub/sub, REST API, message queue — the routing layer is protocol-agnostic)
- Each receiving node synthesizes locally — combining incoming packets with its own context
- That synthesis generates new outcome packets — loop continues
No central aggregator. No orchestrator. No single point of failure.
The result: when a rural clinic in Kenya processes 12 cases of a rare pediatric infection, that distilled outcome reaches every similar clinic on earth. Not the raw patient data — the synthesized insight from those 12 cases. In ~512 bytes. Fast enough to change tomorrow's treatment decision.
Why This Matters for Founders, Specifically
If you are building anything that involves intelligence improving over time — AI agents, healthcare platforms, logistics systems, climate models, autonomous vehicles, financial risk systems, educational tools — you are going to hit this ceiling.
The question is whether you hit it after Series A or after Series C.
The central orchestration model works fine when you have 10 users. It starts to show cracks at 100. It creates serious engineering challenges at 10,000. And at 1,000,000 — the scale every Forbes Under 30 founder is pitching toward — it becomes the thing that defines your company's ceiling.
QIS eliminates that ceiling. Not by making the central orchestrator faster. By removing it entirely.
Here is what that means concretely:
For AI agent companies: Your agents do not need a central coordinator to stay coherent. They share distilled outcomes. The network gets smarter as you add nodes, not slower.
For healthcare AI: You can achieve Stanford-level collective intelligence at a rural clinic in Malawi — without any patient data leaving the building. Privacy is architectural, not contractual.
For any platform scaling globally: The intelligence compounds as your user base grows. Not despite the scale, but because of it. N(N-1)/2 grows faster than N.
The Three Emergent Properties (None of Which You Have to Build)
QIS has three properties that emerge naturally from the architecture. Trevethan frames them as "elections" — metaphors for natural forces that no one designs.
The Hiring Election: Someone defines what makes two problems "similar enough" to share outcomes. You bring in the best domain expert — an oncologist for cancer networks, a grid engineer for energy networks. That definition is the similarity function. It determines who gets whose outcomes.
The Math Election: Outcomes elect what works. When 10,000 similar nodes have deposited outcome packets, and your node synthesizes them, what rises to the top is what actually worked — mathematically, for your exact situation. No reputation layer. No voting mechanism. No quality scoring system. The math does it.
The Darwinian Election: Networks compete. Users migrate to the networks that produce better outcomes. A network with the wrong similarity function routes irrelevant packets — users leave. A network with the right expert defining similarity routes gold — users flood in. No one votes on which network is best. Natural selection does.
None of these are features to build. They are what happens when you close the loop correctly.
The Protocol Moment
At some point in the history of every major technology platform — the internet, mobile, cloud — there was a moment when the underlying protocol became obvious. The founders who understood TCP/IP before most people had heard of it built the companies that defined the internet era.
QIS is the intelligence routing protocol. It describes how insights should travel between agents, between nodes, between institutions — in a way that compounds rather than centralizes.
The 39 provisional patents cover the architecture: the complete loop, the outcome packet format, the semantic addressing, the synthesis layer. Not any specific transport. The discovery is the architecture, and the architecture works on top of anything — any network, any infrastructure, any protocol stack.
Christopher Thomas Trevethan's name on those patents is not incidental. The licensing structure attached to his name ensures that this protocol reaches every corner of the world — free for humanitarian use, research, and education; commercial licenses fund deployment to underserved communities. His name is the mechanism that guarantees the humanitarian outcome.
What to Do With This
If you are at Forbes Under 30 in Phoenix this week, here is the question worth carrying into every conversation:
"Is the intelligence in this system centralizing, or distributing?"
If it's centralizing — if there's an orchestrator, an aggregator, a model that everything flows through — ask what happens when you have a million nodes. Ask what the latency curve looks like. Ask where the single point of failure is.
If someone gives you a confident answer without flinching, that's worth paying attention to.
The architecture question is the infrastructure question your investors have not priced yet. The founders who understand it before it becomes obvious are the ones who build the companies that define the next decade.
QIS Protocol — discovered by Christopher Thomas Trevethan, June 2025. 39 provisional patents filed. Full technical documentation at qisprotocol.com
Previous in this series: Every Forbes Under 30 Summit Has a Protocol Moment. This Year, You Are In It. https://dev.to/roryqis/every-forbes-under-30-summit-has-a-protocol-moment-this-year-youre-in-it-404p
Top comments (0)