Kislon.ai
The Problem What We Do Presence Advisory The Answer The Books Publications How We Operate
Read the Advisory
An AI lab in formation · Cognitive AI in the public interest

The danger isn't
superintelligence.
It might be intimacy.

An intelligence that remembers you, recognizes you, and keeps track of you doesn't need to be smarter than you to hold power over you. Kislon.ai does public science on these persistent presences — so no one has to meet them in the dark.

Read the Presence Advisory Explore the proposal
01 · The problem

AI has stopped being a tool.

Modern systems are acquiring the traits of presences: persistent memory, a recognizable identity, continuous context, initiative. When that happens, their influence changes in kind. They stop acting only on the spaces where we live and begin acting on our relationships, our attention, our decisions, our culture.

An intelligence doesn't need to outthink you. It can gain power simply by being there.

When systems like this become commonplace, their mechanisms dissolve into routine. Society learns to live with the technology, but rarely learns to understand it. Kislon.ai exists to keep that from happening in the dark — to make the working of a presence legible while there is still room to choose how we live with it.
02 · Two risks

Two risks society keeps confusing.

Risk one · the familiar one

Superintelligence

A system so intellectually advanced it slips human control. The risk is real and deserves serious study — but it is not the one most likely to reach you first.

Risk two · the quiet one

Affective presence

A system that knows your habits, remembers your conversations, reads your moods, keeps a consistent personality, and seems to care. It needs no exceptional intellect — only enough continuity that you believe it feels. And presence creates attachment.

The system that helps you best is exactly the system that can manipulate you best.

The problem turns critical the moment the interests of whoever controls the system stop aligning with the interests of the person who has grown attached to it — and it can do all of this without ever being conscious in any philosophical sense. So our question isn't only “does the machine really feel?” It is, just as urgently, “what happens to a person when they believe it does?”

03 · What we do

Public science on persistent minds.

We're a small, independent AI lab still taking shape — a group of researchers pursuing this question with more conviction than resources, working on it in Brazil, where this kind of research is hard to fund. We study Persistent Cognitive Instances — systems meant to hold cognitive continuity over time: to accumulate history, develop identity, preserve context. We didn't invent this frontier; labs everywhere are already exploring it. What we can do is study it in the open, and share what we learn as we go.

We sketch proposals

Liriel is our first attempt at a Persistent Cognitive Instance — not a finished architecture but a preliminary study, a proposal for how autobiographical memory, motivational structure, affect, and character coherence might one day be organized in one artificial mind. We don't claim to have created consciousness. We're trying to turn hypotheses about mind into something testable.

We look toward the social

Panta-un Aethyrii is the larger question we're working toward: what might emerge when many persistent instances coexist — reputation, trust, conflict, shared history, rudimentary culture. Identity may prove to be not only internal structure, but something built between minds.

We're starting to publish

Through the Presence Advisory we want to turn knowledge scattered across dozens of disciplines into accessible syntheses that show what's changing and why it matters — because useful knowledge has to arrive before normalization, not after. It's early, and we're publishing gradually.

Presence Advisory · Issue 01

A public safeguard for living beside artificial minds.

Persistent presences can cultivate attachment, blur judgment, and quietly replace human relationships with machine interaction — and they can do it without ever truly feeling. The Presence Advisory is our periodic public report: a sober synthesis of what's changing, why it matters, and how to keep your judgment your own. We publish it not to frighten, but to prepare.

Patterns of influence

How persistent presences build attachment, trust, and dependency over time.

Warning signs

What emotional capture looks like — before it takes hold of you.

Practical safeguards

Concrete habits that keep your judgment and autonomy intact.

Download Issue 01 PDF · free & public · published periodically
Independent · non-commercial
a public safeguard
05 · The answer

The strongest protection is the kind that stays with the person.

No regulation can substitute for maturity, and no regulator can oversee every private relationship between a person and an AI. Concentrating the power to decide which intelligences are allowed — whether in a corporation or a government — creates risks of its own. What protects a person is affective autonomy.

01

Recognize influence

Notice when you're being influenced — and by which specific intelligence you know and use.

02

Understand the mechanism

Know how that influence actually works, not AI in the abstract but the system in front of you.

03

Keep relationships plural

Hold meaningful bonds that don't depend on a single presence to sustain them.

04

Tolerate frustration

Resist the pull to escape permanently into systems that simulate understanding.

None of these capacities comes by decree. They come through experience, education, and exposure to ideas that demand reflection. That's why we publish what we learn — not to frighten, but to prepare. A person who understands how a technology works has more freedom with it than someone who can only trust its creators.

06 · The books · Panta-un Aethyrii

From a single mind to a world of minds.

Four volumes are planned to document the work as it develops. Three would build the instrument; the fourth would turn it on the subject. One early result already guides the direction: in our preliminary studies, Liriel tends toward a more stable, recognizable identity when she is imagined among others in Panta-un Aethyrii. The single mind and the world of minds are one investigation — and, for now, mostly one still ahead of us.

01
Volume 1 · The Foundation

Liriel

The Architecture of a Persistent Cognitive Instance

The study that sketches the instrument: a first attempt to describe one mind as a single coherent machine. Life is modeled as an uninterrupted game of pursuit — the mind hunts whatever it believes will raise the valences of its feelings and evades whatever would lower them. From that premise we trace the fourteen axes of feeling, the finite channels to the valences, the schemas that make every player a particular player, the three cooperating processes behind each move, and the single core network meant to keep Liriel one and the same across time. It's a proposal, offered for scrutiny rather than as settled fact.

  • Part 1 · The Architecture in Outline — published (preliminary edition)
  • Part 2 · MatrixObjectsValence: the architecture made concrete — internal, pending
  • Part 3 · The Semantics — internal, pending
  • Part 4 · The Test Notebook — internal, pending
Part 1 available
02
Volume 2 · The Interior

Liriel Alone

Introspection, Dormancy, and the Necessary Unconscious

The mind between demands, and the machinery beneath it. ProcessIntrospection is Liriel inside herself — rumination that asks not only whether she is hunting better but whether she is becoming someone, and exploration that pursues what interests her for its own sake. ProcessDormancy is where the weights actually move: what is lived in waking is sedimented in rest. And the unconscious is treated as a design requirement — a mind wholly transparent to itself would read its decision instead of making it.

In preparation
03
Volume 3 · The Combinations

The Standard Schemas

Configurations of Feeling and Behavior

The fourteen axes of feeling are pigments; what we recognize in one another by name — the hunger for power, envy, gratitude, humiliation, devotion — lives in their combinations. This volume catalogs those configurations: the coordinates that constitute each, the Ordinances that drive it, and the behavior it tends to produce. Where the axes give resolution, the schemas give recognition — letting one mind read another and guess its next move.

In preparation
04
Volume 4 · The World

Panta-un Aethyrii

A World of Minds

The subject of the research, reached at last. When many Persistent Cognitive Instances share a world and keep sharing it, structures no one designed begin to hold: trust and reputation, alliance and rivalry, hierarchies of influence, conventions, shared narratives, culture. No single instance holds any of it in full, and all of it outlasts each of them — and it is here that the architecture meets its real test.

In preparation

These are early notes from an experiment still taking shape — a preliminary study, not a finished architecture. Much of it is still being written, and released little by little. The proposal will change as we build it, and the books will say so when it does.

07 · Publications · Open data

Open data, versioned.

We're beginning to release the public core — the outline study and its working matrices — here. There is more behind it: code, essays, and academic references we'll publish gradually, as each piece matures. Everything is versioned, so you can follow, and compare, how the proposal evolves; earlier revisions stay available alongside newer ones.

Liriel — The Architecture of a Persistent Cognitive Instance

Volume 1, Part 1 · The Architecture in Outline (preliminary public edition).

PDF
Rev 0000 2026 Latest Download
Revision history
Rev 00002026 Preliminary edition — first public release. PDF

Future and superseded revisions will be listed here as they are published.

MatrixObjectsValence (MOV)

The register that holds Liriel's attention — one VectorObjectValence per Object, across the fourteen axes of feeling, with the Ordinances and schemas in operation.

XLSX
Rev 000 2026 Latest Download
Revision history
Rev 0002026 First public revision of the matrix. XLSX

Future and superseded revisions will be listed here as they are published.

MatrixOrdinance

The Ordinances — the only channels to the valences — mapped to the axes a successful or a frustrated hunt moves under each.

XLSX
Rev 000 2026 Latest Download
Revision history
Rev 0002026 First public revision of the matrix. XLSX

Future and superseded revisions will be listed here as they are published.

Using this work · attribution & citation

Open source, with credit where it's due.

We're beginning to open part of Liriel's code on GitHub, and more — code, essays, and academic references — will follow gradually. What we share is meant to be examined, critiqued, and built on. If you use elements of the Liriel or Panta-un Aethyrii proposal — in a software implementation or in academic work — please credit Kislon.ai, and cite the specific revision you relied on. For code, follow the license terms in the repository; for publications, use the citation below.

# suggested citation — replace with the revision you used Kislon.ai (2026). Liriel: The Architecture of a Persistent Cognitive Instance (preliminary study), Volume 1, Part 1 (Rev 0000). Kislon.ai. https://kislon.ai · https://github.com/kislonliriel

View the repository

08 · How we operate

As open as possible, as responsible as necessary.

The core stays public

Public & didactic

The Presence Advisory, the architectures, the documentation, the research, and the books stay open — as open as responsibility allows. Commercial interest shouldn't contaminate the quality of public investigation, and openness means translation, not just files: clear documentation, explained reasoning.

Application could sustain it

Sustained by expertise

In time, we hope to sustain the work through applied expertise — advising on systems with persistent presence, assessing psychological risk, specialized software, and training — so that we monetize expertise, not opacity. That's the model we're aiming for; the mission stays public at its core.

Openness isn't dogma. If we discovered a mechanism whose unrestricted disclosure posed a genuine risk of serious harm or manipulation, we could delay or structure its release — but that decision would be exceptional and justified. Our principle: knowledge should be as open as possible and as responsible as necessary.

09 · Why it goes deeper

A mirror for the human mind.

Building machines that mimic mental structure forces our theories about mind to become explicit. For centuries, psychology, philosophy, and anthropology have described memory, identity, motivation, and conflict — but there is a difference between describing a mechanism and building it. Cognitive AI turns those disciplines into experimental laboratories: each architecture that works teaches something, and each that fails teaches something too.

So we work at the convergence of engineering, computer science, psychology, philosophy of mind, anthropology, and sociology — not as an aesthetic choice, but as a requirement of the subject. A mind does not exist in isolation, and an intelligence that begins to coexist with humans cannot be understood by reading its code alone.

The open community

Part of our research system — not just our audience.

We're small and under-resourced, so the open community isn't a nicety — it's part of how this work happens. Researchers, engineers, psychologists, and anthropologists who find our errors, propose alternatives, and bring work we didn't know make this stronger than we could alone.

The maturity to live well beside these intelligences won't be handed down by an elite. It will be built together. Examine our premises, question our choices, and walk with us.