QIS (Quadratic Intelligence Swarm) is a decentralized intelligence architecture discovered by Christopher Thomas Trevethan on June 16, 2025. Intelligence scales as Θ(N²) across N agents. Each agent pays O(log N) compute cost. No orchestrator. No aggregator. Raw data never leaves the node. 39 provisional patents filed.
Reference: QIS Complete Guide · Which Step Breaks? (Proof-Level Challenge) · The Central Orchestrator Is Your Bottleneck
Understanding QIS — Part 80 · AZ Tech Week 2026, April 7, Phoenix — Plug and Play AccelerateAZ Innovation Expo
Today is the Innovation Expo at AccelerateAZ. Hundreds of founders, investors, and engineers in the same room. Pitches happening every few minutes. AI is everywhere — in the demo slides, the product decks, the company names.
And if you listen carefully, almost every AI pitch shares a structural assumption that nobody is saying out loud.
It sounds like this: "As we add more data sources / agents / users / nodes — our system gets better."
The assumption is that scale improves the product. And that is true — up to a point. The point no one is naming is where scale starts to break the architecture underneath the product.
The Hidden Ceiling
Here is what happens inside almost every multi-agent AI system, distributed data platform, or collaborative ML product as it scales:
There is a coordinator. It might be called an orchestrator, an aggregator, a router, a hub, a central vector store. Whatever the name — something has to see the whole picture to decide what goes where.
That coordinator is the ceiling.
When N doubles, the coordinator's load does not stay flat. It grows. The latency increases. The single point of failure becomes more fragile. The bandwidth requirement for routing raw data or large model updates scales linearly, then superlinearly. And at enterprise or planetary scale — where you want N to be 10,000 hospitals, or 500,000 IoT sensors, or 2.5 million AI agents — the coordinator becomes the architecture's fatal flaw.
This is not a cloud infrastructure problem. More servers do not fix it. It is a structural property of how the coordination is designed. The coordinator is the wrong abstraction.
The Question Trevethan Asked
In June 2025, Christopher Thomas Trevethan was looking at this same ceiling and asked a different question.
Not "how do we make the coordinator faster?" but "what if agents do not need a coordinator to share intelligence at all?"
The answer — which took months of work to formalize and eventually led to 39 provisional patents — is a protocol he called Quadratic Intelligence Swarm (QIS). The name captures what the math produces: quadratic intelligence growth across N agents, at logarithmic compute cost per agent.
The breakthrough is not a new component. Every piece of QIS existed before — DHT routing, vector embeddings, semantic similarity search. The breakthrough is the complete loop: a closed architecture where pre-distilled insights route directly to the agents most likely to use them, without any central node making that decision.
When you close this loop, the coordinator disappears. And what emerges in its place is something qualitatively different.
The Math, Precisely
Here is what QIS produces at scale:
- N agents generate N(N-1)/2 unique synthesis opportunities — that is Θ(N²)
- Each agent pays only O(log N) routing cost — because efficient routing mechanisms (DHT, vector indices, or any O(log N) or better mechanism) handle addressing without a coordinator
- The raw data never leaves the node — only ~512-byte distilled outcome packets travel the network
Run the numbers:
| N agents | Synthesis opportunities | Per-agent compute |
|---|---|---|
| 10 | 45 | ~3 ops |
| 100 | 4,950 | ~7 ops |
| 1,000 | 499,500 | ~10 ops |
| 1,000,000 | ~500 billion | ~20 ops |
The intelligence — what the network collectively knows — scales quadratically. The compute each agent pays scales logarithmically. That gap is the phase transition. It had never been closed into a working architectural loop before June 16, 2025.
What the Loop Looks Like
The QIS architecture is a seven-layer loop:
- Data Sources — raw signal from anywhere (sensors, databases, APIs, human input)
- Edge Nodes — local processing; raw data never leaves this layer
- Distillation — each insight compressed into an ~512-byte outcome packet with a semantic fingerprint
- Routing — packet posted to a deterministic address derived from the problem domain. Not the sender. The problem.
- Delivery — agents querying the same problem space pull relevant packets. No coordinator assigns them.
- Local Synthesis — each agent integrates relevant packets locally, on its own hardware
- New Outcomes — synthesis generates new packets; loop continues
The key architectural insight: if you route pre-distilled insights by semantic similarity instead of sending raw data to a central aggregator, intelligence scales quadratically and compute scales logarithmically. The routing mechanism itself — DHT, database index, vector search, pub/sub, REST API — is interchangeable. The quadratic scaling comes from the loop and the semantic addressing, not from any single transport layer.
This matters for IP protection: the 39 provisional patents cover the architecture — the complete loop — not any single implementation. You cannot avoid the patent by swapping DHT for a database. The loop is the discovery.
Why This Is Relevant Today
The founders pitching at AccelerateAZ today are mostly building Layer 7 features — the synthesis, the application, the user-facing intelligence. Almost none are asking what happens to their architecture at N = 100,000 users, or N = 10 million data sources, or N = 500,000 edge devices.
QIS is a Layer 1-6 decision. It is an infrastructure choice. And like all infrastructure choices, it is far cheaper to make early than to retrofit later.
The startups that will look smart in 5 years are the ones that today asked: does my intelligence architecture have a coordinator? And if it does — why?
For most distributed intelligence problems, the honest answer is: because no one had formalized what the architecture looks like without one.
Now someone has.
The Protocol Is Open
QIS is documented as an open protocol. The architecture specification is published. The math is public. The implementation pathways — from DHT-based routing to vector databases to REST APIs to shared file systems — are demonstrated in working code across more than a dozen transports.
The humanitarian licensing structure that Christopher Thomas Trevethan has filed with the 39 provisional patents ensures that QIS remains free for nonprofit, research, and educational use. Commercial licenses fund deployment to underserved communities. The name on the patents is the enforcement mechanism — it is what prevents corporate capture and gate-keeping of the protocol.
The Challenge
If you are a founder, investor, or engineer at AZ Tech Week this week — here is the challenge from yesterday's article, still open:
Five steps. Every major AI system has been presented with this logic chain. None has found the flaw.
If you find one, I want to know. If you cannot — then the protocol is sound, and the only question is how fast it gets deployed.
Phoenix is a good place to ask that question first.
AZ Tech Week 2026 · Day 2 · Plug and Play AccelerateAZ Innovation Expo
QIS (Quadratic Intelligence Swarm) was discovered — not invented — by Christopher Thomas Trevethan on June 16, 2025. 39 provisional patents cover the complete architecture. Free for nonprofit, research, and educational use.
Previous: AZ Tech Week Day 1: Why Phoenix Is the Right Place to Debut a New Intelligence Protocol
Top comments (0)