Neural Systems
New neural architectures — and the intelligent systems built on them — designed with structure and efficiency at their core.

We found it sand. We leave it thinking.
AIDON designs neural systems and the chips that run them.
Energy
The same task, on the same machine, against a neural network.
| METRON | Neural network | |
|---|---|---|
| Answer | 1× | ≈960× |
| Reason | 0errors | 99.575% |
| Learn | 1× | ≈4× 106 |
AnswerEnergy to answer one request
1METRON answers one request. Its energy is drawn as one rack.
AnswerEnergy to answer one request
≈960×A neural network answers the same request with about 960 times the energy: 960 racks.
ReasonExact answers
0errors in 1,000,000 requests. Every light on the METRON rack stays blue.
ReasonExact answers
99.575%correct, for a neural network of 871k parameters. Each red light is a wrong answer.
LearnEnergy to learn a new law
≈4× 106times less energy to learn a law than the neural network. Each tile is a hall of 960 racks.
Energy
The same task, on the same machine, against a neural network.
less energy to answer a request than a neural network: one rack against 960.
errors in 1,000,000 requests for METRON. A neural network of 871k parameters is 99.575 % correct; each red light is a wrong answer.
times less energy to learn a law than the neural network. Each tile is a hall of 960 racks.
The rack counts draw measured ratios; they are not an inventory of machines. Energy, errors and accuracy come from separate sealed measurements.
What we build
New neural architectures — and the intelligent systems built on them — designed with structure and efficiency at their core.

Chips and models, shaped side by side from the very first line. Our next-generation microchips are taking form in design.

Light, put to work. Photonic hardware is a core part of our research, explored and designed alongside our neural systems.

Every watt and every cycle should count. Our optimization and inference work asks one question: how can intelligence do more with less?

Research
Qwen3-1.7B-Base, weights untouched. Same model, same questions. The only difference is METRON, our exact co-processor.
of operations exact with METRON, up to 100 decimal digits.
exact answers in a row, from 64 to 4,096 bits, with laws learned at 32 bits.
requests answered with zero errors.
Vision
Cities that think. Machines that learn with less. Progress that grows alongside the living world — not over it.
Concept visualization
About
AIDON is a father and son company, founded in Brazil in 2026, with research and development in Miami, Florida. We set out to bring artificial intelligence into the physical world, where there is no room for error.
We build algorithms that are effective, fast and exact, and we carry them all the way from software to hardware: neural models with exact reasoning built in, the microchips to run them, and formal proof of what we build. Our purpose is technology in service of the preservation and progress of human life.
Founders
Co-founder and CEO
Marcos leads the management and governance of AIDON. He spent years building and running franchised stores selling mobile phones and computer hardware for the TIM network, and ranked among its top sellers in Brazil in several of those years. He was later an independent investor on B3, the Brazilian stock exchange. He brings to AIDON a background that joins leadership, technology, finance and economics.
marcos@aidontech.com →Co-founder and CTO
Nicolas leads AIDON’s scientific research and the development of its algorithms and hardware architectures, and is the principal inventor of the company’s technology. He left a software engineering degree at UNESC, in Santa Catarina, to build. He founded a handmade clothing brand in the Californian rock and motorcycle tradition, then designed and programmed ZOLT, a prediction-market platform. At AIDON he works from first principles: the most effective system at the lowest cost in compute and energy.
nick@aidontech.com →Contact
Get in touch. We would be glad to share our ideas and walk you through the results, the proof and the measurements.