01
Fast and slow
An instinct answers at once; a reflection takes its time. The core listens to both before it acts.
OUR MODELS / A CORE THAT DOES NOT SPEAK
A core that does not speak.
The Cristallo model is not another chatbot. A fast instinct and a slower reflection work together and weigh the for and against of each action. When it needs words, it uses today's models through a channel of its own.
Instinct →Reflection →Decision →Words
02
Words are borrowed
Language comes from today's models, through a channel of its own. The core keeps the judgement.
03
Night regenerates
At night it reviews, dreams and unblocks: the day's traces become tomorrow's footing.
The core and its two rhythms
A closer look
Why a model that does not speak?
What we grow
Most models are judged by what they say. Ours is grown for what it decides: whether to act, wait, ask, or think longer.
Speaking well is a different skill, and others already do it very well. We borrow it when we need it.
What is the workshop?
How it learns
Our days leave traces. In the workshop, the best of them become examples, read by several models before they are kept.
A small model learns from them slowly, with witnesses that check it did not forget what it already knew.
Trajectory
A direction we are growing — not a finished result.
A model that weighs before it speaks.
Instinct and reflection, side by side.
Words through a channel of its own.
Translation is not knowledge of the core: it is a vein towards other resources.
Traces become examples.
Seeds read by several models, kept with witnesses.
A small model of our own, grown from what we lived.
It will be judged by its decisions, in the rooms, beside the others.
