Dialogic and collaborative intelligence
COGI.CO ERGO SUMUS
We think together, therefore we are.
COGI.CO is the deliberative organ through which an organisation thinks with many voices, remembers, chooses among measured alternatives and answers for what it says. The method is human and comes before the models; the models are guests, and replaceable.
Application scenario
The rule that comes before everything
First, who is here. Then, who is missing. Always, who decides.
When a real person can be involved, they are involved. The selftwin comes afterwards, with a precise task: to give voice to those who are missing, to widen what the dialogue can see, to return positions that the people present can take up, contest and correct.
The selftwin complements. The person decides.
Every sentence about COGI.CO is read with one question: who stays at the centre? If the answer is the technology, the sentence gets rewritten.
What COGI.CO is
An organisation receives many answers every day. It needs its own.
COGI.CO is where the organisation forms that answer: it convenes the voices that make it up — people who are present and, for those who are missing, selftwins with a mandate — brings them into dialogue under a protocol, measures where they converge and where they diverge, keeps what was said and by whom, and hands to those who hold the mandate a recognisable position they can answer for.
Language models take part as guests: they enter a lane, with a certificate, within a mandate, and can be replaced without the organisation losing memory, identity or responsibility.
Four things COGI.CO leaves to others
It proves itself in acceptance testing, with the network switched off.
Selftwin
The selftwin has standing.
It is the computable unit of representation: it gives voice to those who have a right to speak and are missing, with a signed mandate, a source for every line and a right of reply. It formulates, it never decides. No synthetic clone of a real person.
It represents domains: a group, a population, a process, an organisation, a thing, a rule. People appear as referents who sign, reply and recalibrate.
Three constitutive constraints of the DNA
Pertinence is the one constraint the selftwin checks by itself at every question — “is this mine?” — and it makes the other two enforceable.
The agent acts, the selftwin represents. The second authorises the first.
How a selftwin is formed: PLÁSIS, in eight phases
The face is fungible, the identity is not. Text, voice, avatar, embodiment: every face declares itself; no face is the face of a real person.
THÝRA
The organisation has an organ of its own.
THÝRA is the organisation's selftwin: it composes the voices, measures divergence, keeps the registry. It represents, remembers, chooses among measured alternatives and acts under a signed mandate: it formulates for those who decide, with a scale of guarantee and an executable mandate.
The ladder of modules, written for the hospital: you enter through the process, rise to the organ; the network is the next contract.
The entry product
First test bed: Italian healthcare. The axes are written for every sector — industry, schools, municipalities, railways, banking, courts, newsrooms, museums, energy communities, civil protection — and in four languages; every scenario is rewritten in the system that adopts it.
VALUE+
The method that moves everything.
VALUE+ is the method by which change is generated by those who live it. Imposed change produces box-ticking; change generated by the people who inhabit it produces value that lasts. It precedes COGI.CO and is its practical root.
One cycle: VALUE+ and the life of the selftwin
The dialogic session is the human form of the convergence proof.
The certified conductor leads the session with the VALUE+ method, calibrates the selftwins and prepares the dashboard. Certification includes VALUE+ as the method for conducting sessions among people: whoever calibrates a selftwin must first know how to lead a session among people.
Application scenarios
Who is missing from the room? Twelve situations, told as stories.
Each scenario starts from a situation the reader recognises, says who is there and who is missing, and shows what changes when the absent voices come in with a mandate. Each scenario is a hypothesis, declared as such: the proof is a pilot, with five measures before and after. Every population selftwin is formed on archetypes and aggregated data, never on identified people; where the group is small, the selftwin stays silent.
And your situation?
In this situation, how could it be useful to me?
Describe your case in your own words: who decides, who is involved, what stays stuck. The guide formulates a first hypothesis according to the method; a person reads it over with you.
Simulation
How COGI.CO acts, layer by layer, in a situation that is both human and artificial.
A demonstration session, step by step: who speaks, which layer of the wafer lights up, what enters the registry and how divergence changes. People, selftwins, guest models and agents have distinct roles, and the simulation keeps them distinct.
The wafer: eight layers, people on top
The room
The session
EUTHÝNA registry
Demonstration with example values; module names follow corpus v11.1. No real data, no real person.
The nine axes
Nine axes that hold each other.
Remove one and the other eight weaken. This mutual dependence is why COGI.CO is a system before it is a catalogue of functions.
The offer
What is sold is the method in operation.
Measures, registries, mandates, calibrations, conductors. Priced per organ, perimeter and capacity; model consumption is passed through at cost, separately.
Return is counted on the three planes of the Return axis, and only what is prudent to count enters the cash line. The count leaves jobs out: freed capacity is time returned to people, and the tasks people learn by stay with those who learn. Nobody guarantees the outcome: the method is guaranteed.
Proof
What you can download, upload and verify.
The reference core, the executed acceptance tests, the founding documents and the tools. Everything claimed here points to an axis and a gate.
The Schrödinger's cat sandbox
What is in the box, and why the cat.
What a sandbox is
A sandbox is a child's sandpit: an enclosed space where something is tried without consequences outside. In software it is the isolated environment where code runs without touching the rest; in the AI Act it is the supervised space where an innovative system is developed and tested under the competent authority's guidance, before it is placed on the market. Here it is the area of the site where COGI.CO puts itself to the test: the core runs without a network, and what you upload stays in your browser.
Why the cat
In Schrödinger's box the cat stays alive and dead at once until someone opens it and looks. An application hypothesis sits in the same state: confirmed and refuted together, until a pilot puts it to the test. The sandbox is where the box opens, and what decides the look is five measures, before and after. A pilot that can fail is what makes it a proof; a proof that can only succeed is a promise.
Hypothesis in the box
·
The box opens with the measures.
Demonstration with example values from the THÝRA Flow reference case; the real numbers come from the pilot, and they can refute the hypothesis.
What is inside
Terms of use are declared in each package. The code uses the standard library only and runs with the network switched off.
Coming soon
Bring your case.
A corpus for a convergence proof, a selftwin to certify, a rule to adopt. The SHA-256 fingerprint will be computed in your browser, before anything is sent.
Uploading is under construction and not yet active. To bring a case, write to us through the contact form.
Drop files here, or choose.
Protocol v11.1-2026-09-08 · acceptance re-run on 1 October 2026 on thyra_core.py, standard library only, network switched off, on synthetic data: the core is an executable specification; the field proof comes from the pilot.
The radar of versions
Self-assessment on twelve axes, scale 1–5. v11.1 is the proposed score (open item A46), to be confirmed at the ninety-day re-scoring.
Agora Coming soon
The forum for conductors, referents and anyone who wants to understand.
Here positions carry a signature and a source. Design preview: the conversations are examples written to show how it works.
Dialogue
Talk with the COGI.CO guide.
A selftwin that declares itself: it speaks for COGI.CO under a mandate from management, answers within its domain and says so outside it. It formulates; people decide. In this prototype the answers are written examples.
Contact
Start the dialogue.
A person answers before any representation. Write to hello@cogi.co, or ask for a convergence proof for your organisation.
hello@cogi.co
COGI.CO was born in Florence from twenty years of work inside complex systems, beginning with healthcare: an experience of systems that went on to give itself a technology. The method is human; the axes are written for every sector and every language.
A minimal lexicon for newcomers