Mindscan Labs · Frontier Tech Research

AI·Quantum Computing·Neuroscience

Any researcher should constantly ask the same question: what if?
This is how ideas are born, discoveries happen, and humanity and civilization change.
It takes creativity and empathy to get there.

Mindscan Labs is an independent lab at the frontier of technology.
It is not afraid to try the boldest ideas, and it tests them the scientific way: hypotheses stated before the test, data never seen in development, results anyone can check.
What holds up becomes a product.

Products

NAI-fnet

A neural network made of formulas, compiled for the chip

  • Grows a model one neuron at a time; every neuron is a short formula
  • Plain C for a microcontroller: no runtime, no dynamic memory
  • Kilobytes of code, a fraction of a microsecond per prediction
  • The C code matches the model bit for bit, with an input range monitor
  • Ten registered tests: robot arm, traction motor, wind farms, airliner, tables
NAI-fnet

NAI

No-Ansatz Inference: finds the law behind your data

  • A system that discovers laws in data
  • The result is a formula you can read and verify
  • No hidden templates, no black box, no language model
  • Runs locally on a consumer laptop CPU
  • 1st on SRBench, SLDBench and LLM-SRBench against the results they publish
NAI

NAI Whisper

Compressed OpenAI's Whisper large-v3-turbo

  • 3 times faster, 3 times smaller
  • The same quality as the original
  • Runs locally on a consumer laptop CPU
  • Own interpreter, no dependencies
  • Compressed with NAI
GITHUB

ConUI

A terminal-like UI framework and design system

  • C++20, OpenGL, GLFW
  • Ready to build
  • Playground demo app
  • Theme editor
  • Claude skill included
GITHUB

Research

Virtual sensors from networks of small formulas: eight registered tests of NAI-fnet

Preprint, November 2026

doi.org/10.5281/zenodo.24163946

The whole f–I curve of a neuron from its rheobase and one current step

A pre-registered test of a one-parameter logarithmic law on the Allen Cell Types Database. Preprint, October 2026

doi.org/10.5281/zenodo.23582787

Track record

Benchmark reports, registered tests with their fingerprints, and accepted contributions to the OEIS, the On-Line Encyclopedia of Integer Sequences

github.com/mindscanlabs/NAI-track-record

Contacts

lab@mindscan.org