Long-term memory for AI · Observe · Remember · Dream · Convene

THE
INTELLIGENCE
IS RENTED.
THE MEMORY
IS YOURS.

You've explained the same constraint five times. You switch AI tools and start over. Every preference, every piece of context, every thing your last AI finally understood — gone. That feeling has a cause. The cause has a shape.

Read the Substrate

Scroll

The Gap

The layer
your AI was
never given.

Your AI has a context window. It is sharp, expensive, and short. When the conversation ends, the window closes, and the next conversation starts from nothing. That is short-term memory — what your AI thinks with. It is not what your AI remembers from.

To paper over the forgetting, the industry built RAG: a vector database in deep storage. It is the basement archive, full of raw tape. Slow. Expensive. Wrong, sometimes badly. The giants have spent two years and billions of dollars trying to make it feel like memory. It doesn't.

The thing your AI is missing isn't a feature. It's a layer. The layer between the context window and the archive. The layer that consolidates. The layer that prepares. The layer that already knows what mattered before you ask.

You have that layer. You use it every night. Your subconscious runs deterministic processes on the day's experience while you sleep. Themes surface. What repeats strengthens. By morning, recall is prepared.

Your AI doesn't have a subconscious. MOOTx01 is that subconscious.

How It Works

Four behaviors.
One system.

Observe

Captures every conversation, idea, or fact you make it aware of verbatim — in the words it happened in, without paraphrase, without silent rewriting. What you said stays said. The raw signal of your thinking life, preserved exactly.

Remember

Returns ranked, filtered, theme-aware signal. Not chaos from the basement archive — ordered thought, already prepared by the layer that ran while you slept. The AI reasons on signal instead of noise.

Dream

Overnight consolidation. The matrix of what-connects-to-what reweighs itself against what actually happened. Themes surface from your week. By morning, tomorrow's answer is already prepared. The dreaming daemon ran while you slept.

Convene

MOOTs can gather — your work MOOT and your home MOOT, your calendar and your project tracker, your estate and your team's. Each brings what it knows. Nothing centralised. Everything authorised by you.

Where It Runs

Three surfaces.
Same substrate.

01 / PERSONAL

Your Life

Personal estate · Family federation

Your AI knows you across every session, every tool, every device. A work MOOT and a home MOOT — discrete, separated by architecture, not by policy. They do not bleed into each other. The work MOOT does not know about the dentist appointment. The home MOOT does not know about the quarterly review.

Your household can convene — share what the calendar needs without merging estates

Any MCP-capable app contributes its own MOOT with your permission

Switch AI tools. Your memory comes with you

Verbatim capture — what you said stays said

02 / FLEET

Your Machines

Computer health · Multi-device learning

An agent monitors system health and drives corrective behaviour across your machines. It reads from the substrate, proposes actions, captures outcomes, and learns from every result without retraining the underlying model. The agent gets smarter overnight because the dreaming daemon reviewed what worked.

System telemetry as ambient signal — CPU, memory, IO, network, security events

Fleet learns across devices — patterns one machine can't see alone

Three-layer separation: substrate · reasoning · action. The LLM proposes; the actuator validates

Every action proposed, every outcome captured, fully auditable

03 / ENTERPRISE

Your Team

MSP · Hierarchical aggregation

Five tiers of compounding cognition: user → company → industry → MSP. Privacy-preserving aggregation flows upward. Advisory queries flow back down, enriched by every tier. A MOOTx01-backed MSP has a learned matrix at every level. Five orders of magnitude more accumulated structure than ticketing-plus-KB.

Sidecar pattern — existing app attaches a MOOT, no rebuild required

Pattern recognition across companies in an industry, without content exposure

Consent-based participation at every tier boundary — per-tier opt-out

Competitive moat lives in the matrices, not in model weights

The Adoption Path

The adoption path. Your existing
application.
A MOOT
Beside it.

The primary way an organization adopts MOOTx01 is by sidecar. An existing app — one already on the market — attaches a MOOT alongside itself and opens that knowledge to the organization through the ARIA MCP server. The app is not rebuilt on the SDK. It gains a MOOT beside it.

Any MCP client — Claude, Claude Code, or any MCP-capable tool — can read from it. An agent should be able to read the source and wire it up with little guidance. A developer should find it just as easy.

The sidecar pattern is the broad adoption path. It is distinct from building a new app on the SDK or shipping the kit libraries. It is what the demonstration apps exist to show.

"The app is not rebuilt on the SDK. It gains a MOOT beside it and exposes that memory over MCP. Ease of this integration is the product feature."

The Name

An old word.
A lost
meaning.
Reclaimed.

Moot is a word from old English. The assembly where a community brought its memory together — witnessed events, sworn oaths, who decided what last winter. The record lived in the gathering. The gathering was the record.

Modern English kept the word but lost the meaning. Moot now means no longer relevant. We are taking the word back, because the older meaning is what memory actually is: observed over time, kept exactly as it was, available across every AI you use.

You are x01. Hex 01. First person. You are the hero of your story, and your MOOT is the gathering of it. There is no MOOTx02. You do not get upgraded to a later version. Every MOOT begins at first person because every memory does.

"Memories observed over time. A gathering of what was seen, kept exactly as it was, ready before you ask."

The Interface

ARIA — Augmented Recall
and Inference Architecture

Grammar

One Noun

Every call in the ARIA language operates on a drawer — the one noun. Nine verbs. Four adjective categories. The grammar is minimal and complete. It is the contract every consumption surface honours.

Protocol

MCP First

ARIA_MCP is the first consumption surface. It wraps a MOOTx01 estate and projects the ARIA surface onto MCP primitives. Any MCP client reads from your estate — Claude, ChatGPT, Gemini, local models, or anything that speaks the Model Context Protocol.

SDK

Native API

Compiled applications link the kit libraries directly — no server, no wire. The same contract, reached three ways: MCP server, Native API, or Embedded in-process. Three surfaces, one grammar.

Ownership

Open Grammar

ARIA is free. The grammar is open. The estate stays private to the user. You do not sell ARIA; you sell MOOTx01. ARIA is the property that makes it portable and ownable across every AI it speaks to.

Platform

Native everywhere.
Full parity.

Apple Ecosystem

Swift Native

macOS 15+ iOS 18+ Apple Silicon Metal GPU CloudKit Sync

Swift 6 strict concurrency throughout. Raw SQLite, no Core Data. Metal framework for on-device GPU compute. Zero external Swift package dependencies in the kits. The full substrate runs natively on every Apple device.

Everyone Else

Rust Parity

PC / Linux x86_64 aarch64 Full Feature Parity Conformance-Gated

Every Swift kit ships a Rust port, conformance-gated against shared test vectors. Full feature parity. Not a wrapper. Not a port that lags. The same algorithms, the same behaviour, the same ARIA grammar — verified against the same reference implementation.

License

Free to build
and share.
Pay to profit.

Personal · Developer

Free forever

Any individual using MOOTx01 for their own life. Any developer building and sharing non-paid work. Make something new and free with it — go forward, no permission needed.

Mobile App

$500 lifetime

Compiling MOOTx01 into a for-sale mobile application. Covers every point release within the major version. The next major is a new purchase.

Corporate Internal

$5,000 lifetime

Internal-facing applications within an organization. Covers every point release within the major. Not for hosting MOOTs as a service for third parties. Support contracts available.

Service Provider

$5K / year

Hosting MOOTs as a service for third parties. The MSP tier. The sidecar pattern, hierarchical aggregation, and multi-tenant operation.

FedRAMP / CMMC

Negotiated

Commercial hardening kept out of the open core. Contact us for regulated-sector and FedRAMP / CMMC-aligned deployments.

The rule

Build and share free: free.
Profit from what you built with it: get a license.

The blocked act is selling what you built with the code — not who you are.

THE INTELLIGENCE IS RENTED. THE MEMORY IS YOURS.

Four behaviors.One system.

Three surfaces.Same substrate.

ARIA — Augmented Recalland Inference Architecture