NOETELICR.A.B.İ.A

Birlikte başlayan. / A shared beginning.

RABIA / PERSONAL AI

The future,
remembers you.

Instead of starting over with every conversation, continue where you left off together.

Rabia is a personal AI in development: to remember decisions in context, explore sources and help with your work only with your permission.

EXAMPLE USE“What did we decide for the presentation last week?”Return to the decision, with the conversation behind it.Developed by Noetelic · Powered by CCOS ↗
Pronounced “Rabia” · R.A.B.İ.AThe model can change. The story stays yours.
01 / SHARED HISTORY02 / PROVENANCE03 / PERMISSION BOUNDARY
NOETELIC / PERSONAL INTELLIGENCELOCAL FIRST · EXPLICIT PERMISSION

WHY RABIA?

Models change. Your context stays yours.

Rabia’s focus: keeping memory, knowledge sources and action permissions independent of any one model.

CHAT AND MEMORY

ChatGPT · Claude

Both offer memory features that can draw on earlier conversation context. Memory is not unique to Rabia.

RUNNING MODELS

Ollama

Helps you run models on your device and connect them to applications. Running a local model and managing personal memory and permissions are different layers.

PERSONAL CONTINUITY · IN DEVELOPMENT

Rabia

We are developing a local-first personal mind that keeps identity, shared history, source records and permissions separate from the replaceable model.

01 / POSSIBILITIES LAB

In the course of a day.
Four moments, together.

Choose a scene. Explore how the context comes together.

Interactive design example. Not connected to the application or your personal data.

RABIA / THINK TOGETHERILLUSTRATIVE SCENARIO

“What did we last decide for the presentation?”

RABIA

We were going to start with a real use case.

You chose to show everyday value first and leave the technical details until the end.

● The answer, alongside the context behind it.

Explore Rabia’s approach in depth ↗

Local-first personal intelligence infrastructure

As AI learns who you are,
control should remain yours.

R.A.B.İ.A is a local-first personal AI application that preserves shared conversation history, lets you correct its understanding of your preferences, and uses tools only within your permissions. Developed by Noetelic.

R.A.B.İ.A / EXAMPLE SCENARIOIllustrative product view
CONTEXT EXPLAINER

Why this response?

EXAMPLE

I preserved your preference for direct answers and placed technical detail after the outcome.

M
Memory usedPrefers the outcome first, followed by technical detail.
User statementConsistent across 14 conversationsConfidence 0.94
Tool boundaryRead-only · Limited to this task
Outcome verified

Illustrative scenario: numbers and states are not live measurements.

01OWNERSHIP
User-owned memory

Personal records are kept outside model weights. Source-linked records have inspection, transfer and deletion paths; acceptance for older archives with missing provenance remains open.

02PROVENANCE
Visible provenance

Every record retains whether it came from a user statement, observation, web source, inference or dream.

03AUTHORITY
Explicit permission

Tools operate only within the approved task, resource and duration; the outcome is verified separately.

04CONTINUITY
Model independence

Identity, shared history and authority remain in CCOS even when the model changes.

ACTUAL APPLICATION / TAKE A CLOSER LOOK

R.A.B.İ.A on the desktop.

How does a file change happen? See the difference first. Then decide.

  1. 01
    See the change

    Original and new content, side by side.

  2. 02
    You approve

    Explicit approval for the text to be written.

  3. 03
    Verify the result

    The file is read back and the result recorded.

27 September 2026 · Step-by-step capture of the actual app using a sample file. Text edited manually; transitions shortened and file paths hidden.

R.A.B.İ.A / WHAT’S NEW

Your computer. Your boundaries.

Three new steps towards permissioned computer interaction.

02 / MIND ATLAS

Not around a model.
Around your context.

Rabia’s approach separates memory, knowledge provenance and permissions from the replaceable model. That is where it differs from a chat box or model runner alone.

Three-dimensional architectural illustration with CCOS layers at the centre, sourced memory on the left, a replaceable model on the right and a permission boundary at the front.
World Model

Brings together entities, events and confidence. The conversation world is derived from the sourced ledger; switching topics preserves the previous thread’s state.

  1. 01Preparation
  2. 02Deliberation
  3. 03World model
  4. 04Validation
  5. 05Integration
A visual explanation grounded in CCOS source code. The 3D scene is illustrative, not a hardware product.
01

What it remembers has an origin.

“You told me”, “I read a source” and “I inferred it” are treated as different types of knowledge.

02

The model does not own your memory.

The model is a language and reasoning component. Identity, history and permissions belong to the personal layer.

03

A suggestion does not become an action by itself.

Tool use is designed around permissions scoped to a resource, duration and operation. The outcome is checked separately.

03 / PATHS INTO THE FUTURE

A horizon beyond
a single screen.

A vision of tomorrow, with a clear view of where we are today.

A / DESKTOP
Actual application screenshot · Desktop

Actual application screenshot · Desktop

Alongside you, where you work.

A desktop experience bringing together memory, research, tasks and permissioned file workflows.

Development version
B / MOBILE
Actual device screenshot · Android

Actual device screenshot · Android

First contact, in your hands.

The Android introduction and personal-space experience is in development, with the aim of carrying shared context across screens.

Development version
C / PERMISSIONED COMPUTER USE
Permissioned file workflow · Private paths obscured

Permissioned file workflow · Private paths obscured

Capability is not permission.

Understanding a selected window, proposing steps and obtaining separate action permissions. General computer control is not yet publicly available.

Infrastructure and controlled tests

04 / A MEASURABLE STEP

When you say stop, a partial answer should not count as a result.

60/60cancellation trials during generation

CPU tests of the selected local model verified that partial answers were not published across repeated trials of two cancellation paths.

03 Oct 2026 · Repeated cancellation trials, not 60 different tasks.Explore the test scope ↗

CONTEXT THAT CONTINUES WITH YOU

Shared history. Clear boundaries. Your own space.

What you discussed, what you prefer and what you authorised are not the same thing. R.A.B.İ.A is being developed to preserve those distinctions through CCOS.

01 / EVIDENCE-AWARE MEMORY

It does not merely remember. It retains how it knows.

User statements, observations, web sources and model inferences remain distinct record types.

Return to a project“What did we last decide for the presentation?” A scenario for finding relevant shared history with its source.Explore the example →

P
Concise and direct answersUser statement
✓
O
Deep work tends to happen at nightBehavior pattern
✓
I
Local operation may be preferredVerification required
✓

02 / PERMISSIONED ACTION

Every tool call begins with an explicit boundary.

Follow the researchDistinguish a discovered link from evidence actually read and used in the answer.Explore the example →

PERMISSION GATE
RESOURCE
Your selected project note
ACTION
Read-only
SCOPE
This task only

03 / PORTABLE IDENTITY

The model changes. Shared history remains.

The language model is a reasoning component; Rabia’s identity, relationships, memory and authority are held in CCOS.

Update your preference“I now prefer short answers.” Explore how a correction can inform later context.Explore the example →

01LOCAL MODELSELECTABLE
02SELECTED PROVIDERPERMISSIONED
∞R.A.B.İ.A IDENTITYPERSISTENT

04 / HONEST UNCERTAINTY

It does not confuse what it knows, infers and does not know.

“I remember” and “I inferred” are not the same level of certainty.
CCOS architectural principle

A CLOSER LOOK

Follow your curiosity.

THE STORY IS STILL BEING WRITTEN.

Be part of what comes next.
Share your idea today.

Rabia is in development. Public downloads are not available yet. You can share ideas, feedback or collaboration proposals.

Talk to Noetelic ↗
Pronounced “Rabia”. R.A.B.İ.A: Relational, Adaptive, Boundaried Intelligence Architecture.

FREQUENTLY ASKED QUESTIONS

Before you get started.

Test conditions and open boundaries →

NOETELIC / R.A.B.İ.A

Personal intelligence should not be rented. It should be yours.

R.A.B.İ.A is in development. Public downloads and early-access registration are not open yet. This page introduces the product approach and current development status.

Noetelic · Ahmet Özdemir

Noetelic is the company owned by Ahmet Özdemir that develops R.A.B.İ.A. R.A.B.İ.A is a personal AI application that preserves shared history, prioritises local operation and acts only with permission.

Preview registration is not open yet