What is the smallest thing
that can think?
We are deriving the foundations of intelligence, mathematically, from first principles.
The premise
Modern AI works.
No one knows why.
Today’s models are trained to reproduce the outputs of intelligence: text, reasoning, behavior. They capture the appearance of thought with extraordinary fidelity. But an imitation of thinking, however faithful, is not the same object as thinking itself.
Scaling this paradigm produces better imitations. It does not produce understanding, in the machine or of it.
We hold a different premise: intelligence is not a statistical artifact. It is a structure, one that can be defined, formalized, and built.
The approach
We begin where the field never did: with a definition.
Using mathematics and formal logic, we are isolating the minimal cognitive primitives: the smallest set of components from which general intelligence can be composed.
Not neurons. Not attention. Primitives.
From these primitives, we are designing a new architecture. Not a descendant of the transformer, but a different kind of object entirely.
We call it the Mind Model.
The path
The path runs through a definition of intelligence itself.
-
I Define now
A benchmark that formally specifies the minimal components of intelligence, and measures, for the first time, whether any system actually possesses them.
-
II Formalize
A mathematical framework composing these primitives into a complete theory of cognitive architecture.
-
III Build
The Mind Model: the first system constructed from the structure of intelligence, rather than trained to imitate its surface.
Now
COGITO
Every benchmark today asks the same question: how well does the model perform?
Ours asks a different one: does it think at all?
COGITO is a formal test of the minimal cognitive components of intelligence: components we can prove are necessary, and measure in any system, human or machine.
It is the foundation of everything we build after it. It is also, we believe, the measurement the field is missing.
First results, soon.
The people
A small team of mathematicians, logicians, and engineers who believe the most important question in AI has not yet been asked properly.
Write to us
We are a seed-stage company. We work quietly, and we publish when things are ready.
If you invest in foundational technology before it is obvious, we would like to hear from you.