Three Facts That Define QIS
Situation = Address
Two nodes with the same problem compute the same address independently, without communicating. The situation determines the routing destination. No lookup table. No coordinator.
The Mailbox Is Already Full
When you open the mailbox at your address, outcome notes from every node that already solved your problem are waiting. No compute happens at query time. You are reading, not asking.
Raw Data Never Moves
Only outcome packets travel — approximately 512 bytes each. One edge, one outcome, one packet. No patient identifiers. No raw data. Only the validated result routes.
The architecture is the invention. Every component below existed for decades. The closed loop that produces quadratic intelligence scaling did not.
The Complete Loop
Observe
Each node collects data from its own local sources. A hospital aggregates treatment records. A tractor ingests soil sensors. A weather station logs atmospheric readings. The data format does not matter — QIS is format-agnostic.
Raw data stays on the originating device. Always.
Distill
The node compresses what it learned into an outcome packet — approximately 512 bytes. This contains a condition code, a treatment code, an outcome delta, a confidence indicator, and a coarsened population tag. One edge, one outcome, one packet.
No patient identifiers. No raw data. The packet IS the insight — what was the problem, what was done, what changed. Every edge transmits the same ~512-byte packet regardless of whether the edge is a patient's phone, a hospital system, or a tractor's sensor array.
Agriculture: crop type + soil class + yield delta + weather context + season
Automotive: road condition + speed context + hazard type + response outcome
Fingerprint
The situation context is encoded as a semantic address. This address is deterministic — two nodes working on the same clinical question, crop problem, or engineering challenge produce the same fingerprint independently, without coordinating.
The address space already exists in every hospital using SNOMED CT, every farm using crop classification codes, every vehicle fleet using standard telemetry schemas. QIS does not require a new vocabulary. It routes using vocabularies that domain practitioners already maintain.
Route
The outcome packet routes to its matching address at O(log N) or better cost. For 10,000 nodes, that is roughly 13 hops. For 1,000,000 nodes, roughly 20 hops. Cost grows logarithmically while intelligence grows quadratically.
QIS is transport-agnostic. The same protocol works over distributed hash tables, REST APIs, databases, pub/sub messaging, shared folders, or SMS. Any transport that can carry 512 bytes can carry a QIS outcome packet.
Synthesize
The receiving node tallies all deposited outcome packets locally. No central aggregator processes the data. Synthesis happens at the edge — vote counting, Bayesian updating, confidence-weighted averaging — using only the outcome packets that arrived at this address.
For 1,000 outcome packets, synthesis takes milliseconds. The result: an evidence-based answer drawn from every node that has ever solved this problem.
Validate
Three emergent properties naturally validate packet quality:
Election 1: Domain experts define what "similar" means — the fingerprint template. Different networks elect different curators. Competition on curation quality.
Election 2: Outcomes vote by being counted. No algorithm decides which treatment is best. The outcomes themselves — aggregated across hundreds of nodes — produce the answer.
Election 3: Users migrate to networks with better results. Networks that save more lives, grow more food, or prevent more failures attract more nodes. Natural selection at network scale.
Act and Re-Deposit
The node acts on the synthesis. That action produces a new outcome. The new outcome distills into a new packet. The packet deposits back at the same address.
The next node that opens this mailbox finds one more note. 312 becomes 313. The network learned from your reading AND your acting. Intelligence compounds with every cycle.
Without the loop closure, QIS is a database. With it, QIS is a learning network. The complete loop — not any single component — is the discovery.
The Scaling Law
N nodes create N(N-1)/2 unique synthesis opportunities. Communication cost per node: O(log N) or better. The gap widens with every node added.
| Nodes | Synthesis Pairs | Cost per Node | Intelligence : Cost |
|---|---|---|---|
| 10 | 45 | ~3 hops | 4:1 |
| 100 | 4,950 | ~7 hops | 50:1 |
| 1,000 | 499,500 | ~10 hops | 499:1 |
| 10,000 | 49,995,000 | ~13 hops | 4,999:1 |
| 100,000 | 4,999,950,000 | ~17 hops | 49,999:1 |
What QIS Is Not
Where It Works
QIS applies to any domain where outcomes are aggregatable and similarity is definable. The architecture is identical in every case — change three strings (situation description, address fields, outcome format) and the same protocol runs.
Healthcare
Treatment outcomes across hospitals
Agriculture
Yield data across farms and climates
Autonomous Vehicles
Hazard patterns across fleets
Climate Science
Sensor synthesis across stations
Drug Safety
Adverse event signals across populations
Radio Astronomy
Observation correlation across telescopes
Supply Chain
Disruption signals across networks
Rare Disease
Small-cohort outcomes across institutions
39 provisional patent applications filed across 28 domains. The math does not change between them.
Dive Deeper
See the Core Loop
Interactive demo — watch the full QIS loop in 90 seconds
Full Architecture Diagram
All 7 layers from data sources to collective intelligence
Compare Approaches
QIS vs. federated learning, centralized AI, edge AI, and more
Every Component Exists Today
No new science needed — novel composition of proven technology
The Scaling Law
Why N(N-1)/2 intelligence with O(log N) or better cost changes everything
All Articles
78+ articles covering every aspect of QIS Protocol