---
title: "Orbit: A Living Model of Your Life, Built From Your Messages"
description: "Most AI assistants search your data at question time and hope the right fragments surface. Orbit maintains state — a living, continuously updated model of your life. Here's how it works, from orbits to the daily brief."
date: 2026-08-26
author: Bill Simmons
url: https://orbit.me/learn/a-living-model-of-your-life
---

# Orbit: A Living Model of Your Life, Built From Your Messages

The structure of your life — what's pending, who matters, what was promised — exists only *latently*, scattered across tens of thousands of messages nobody has time to re-read. Orbit makes that structure **explicit**: a continuously updated, fully auditable model of your life, refreshed as the messages arrive, and reflected in every surface you use.

Most AI assistants *retrieve*: they search your data at question time and hope the right fragments surface. Orbit *maintains state*. That difference — a living model with memory, corrections, and sources you can check, instead of a search index with a chat window — is the whole architecture.

This article is about how Orbit.me works. I'll walk through the shape of the system — inputs, model, and outputs — then what orbits are and why we organize your life into them, how the model stays current, what it means to accumulate understanding over time, and how every surface you use — including the AI of your choosing — is a view of one model.

And because the model is built from history you already have, Orbit is useful the first morning you connect it. It compounds from there: every message, every correction, every day of history makes it better.

## The shape: inputs, model, and outputs

Start with the inputs: connectors for email, messaging, and calendars bring in your communication as it happens — a set that grows over time. In the center sits one model per person — not a search index, but a set of organized, structured stores under a shared change ledger. And then the outputs: every product surface is a *view* of that one model, and none of them holds state of its own.

![Diagram of Orbit's architecture: input connectors for email, messaging, calendars, and contacts flow into one living model per person — orbits and topic cards, narrative memory, people, events, preferences, and to-dos over an append-only change ledger — which is projected out into the feed, daily brief, chat, calendar feed, personal CRM, and MCP connector.](/images/articles/living-model-shape.svg "The middle is the product. The surfaces hold no state and make no decisions of their own — web and mobile render the same settled answer, and the MCP connector (beta) hands that same model to any AI you choose. Both edges are designed to grow: new inputs enrich the one model, and new outputs are just new views of it.")

## Orbits: the themes of your life

Life doesn't arrive organized. It arrives as a flood. In our studies of Orbit users, people receive 50,000 to 110,000 words a day across their inboxes and messages. At a typical reading pace of 140 words per minute, that's six to thirteen hours of solid reading every day — up to nine novels' worth of text a week — and it arrives interwoven and out of order: a school email between two client threads, a contractor's text in the middle of a job negotiation. Making sense of it demands a context switch at every message. Nobody can hold that in their head.

So Orbit reorganizes it. An **orbit** is a theme of your life: your family, a client engagement, a home renovation, a job search, an aging parent. Every incoming message is routed to the theme it belongs to, and everything Orbit understands about that theme collects in one place.

The point of orbits is **context**. Related topics collapse into a single frame, so you can hold a whole theme in your head at a glance — and the AI can hold it in working memory the same way — instead of either of you reconstructing it from a thousand scattered threads.

When Orbit writes your brief, answers a question, or decides what deserves your attention, it reasons over one theme's context, not the whole archive at once. That's why a brief from Orbit reads like it was written by someone who knows the whole story — not like a list of unread mail.

Inside an orbit live its **topic cards** — the active threads of the theme, each stating what's happening and what it asks of you, and each citing the exact messages it came from — and a **narrative**: the running story of the theme, the connective tissue no single message states.

![Diagram zooming into one orbit: from a set of orbits — Family & school, Client: Acme project, Home renovation, Job search — into a single orbit's contents: topic cards, a narrative, and connections into the stores shared across your whole life.](/images/articles/living-model-orbits.svg "A zoom into one orbit. An orbit is a bounded frame of context: small enough for you to scan and for the AI to reason over, complete enough to stand on its own.")

From there, orbits connect into the stores that span your whole life: the people and organizations involved, the events (declared on a calendar or discovered in conversation), the to-dos, and your preferences and standing rules.

Orbits are also how attention is spent. Most of what arrives in an inbox is bulk — newsletters, promotions, receipts — and a model that studies everything equally becomes a model of your subscriptions, not your life. A learned sense of relevance decides which messages deserve deep analysis, and orbit routing puts the results where they belong.

Underneath it all, the model is **explicit, not a black box**. There are embeddings in the machinery — every message carries a vector for search, and each orbit holds a routing signature — but the model itself is a set of structured records a person could read: every field means something, and every change has an author.

## Freshness: how the model stays current

Orbit updates in near real time. New messages are captured and encrypted within moments of arriving, then summarized and routed to the theme they belong to — and each new signal triggers an update to the model.

Those updates are written as layered changes over the record, never overwrites, and every surface merges the latest changes as it reads. What you see reflects a change the moment it happens, and your own edits always take priority. The answer you get at 9:00 includes the message that arrived at 8:58.

The fast path is paired with two slower ones that keep the model tidy and honest. Each day, a consolidation folds the accumulated changes — the model's and yours — into clean, settled state. Each week, a deeper pass goes back to the original messages and corrects any drift.

![Diagram of Orbit's update loops as a timeline: capture and protect immediately, enrich and route moments later, update on every new signal, reconcile daily, re-ground weekly — and every surface shows updates the moment they land.](/images/articles/living-model-freshness.svg "Fast loops make new information visible right away; slower loops continuously reconcile and correct the model against the original messages.")

## State: understanding, accumulated

The deepest difference between Orbit and an AI that searches your mail is this: **Orbit accumulates understanding**. A search index answers each question fresh, from whatever documents happen to match, and forgets that it answered. Orbit carries its understanding forward — what matters, who matters, and what needs your attention — and revises it as each new message arrives, instead of rebuilding it from scratch.

An example makes it concrete. An interview invitation arrives, and Orbit records the commitment: the date, the people, the preparation it implies. The recruiter reschedules, and Orbit updates that same commitment rather than minting a second copy beside it. You finish the prep and check it off; Orbit remembers.

When you later ask *"where do things stand with Acme?"*, the answer draws on the accumulated state of the whole arc — what was promised, what changed, what's still open — not on a fresh keyword search across a pile of similar-looking emails.

Accumulated understanding also changes what you can ask. A search index answers questions about documents — the ones with a keyword to match. A living model answers questions about your life, including the ones with no keyword at all: *What am I dropping? What needs me today? Who haven't I gotten back to?* Orbit doesn't search for those answers; it reads them off the model it has been keeping all along.

We engineered the model to stay correct as it accumulates. Every conclusion cites the messages it came from, so you can always see why Orbit believes what it believes. Every change lands in a ledger you can open on any card — what changed, and why. And your word is final: correct a name, a date, or a priority, and the correction sticks.

## Projections: one model, many views

The outputs in the first diagram — your personal knowledge — are all views of the same model, never separate products with their own state. And because there is a model, Orbit doesn't have to wait to be asked. An assistant that only retrieves can only respond; Orbit can start the conversation.

The daily brief and radar tell you what changed and what needs your attention; the digest email delivers the same on your schedule. The feed of topic cards shows what's live across your themes. Chat lets you ask and act. The calendar feed publishes the events Orbit caught that your calendar never had. Your people page is a personal CRM holding the living history of each relationship, learned from your actual conversations.

Add a new input and every view gets richer; build a new view and it inherits the whole model on day one.

The **MCP connector (beta)** extends the idea beyond Orbit's own surfaces: it attaches a real-time, continuously updated model of your life to the AI of your choosing. That is different from — and stronger than — the built-in memory of today's assistants, which knows only what has come up in its own conversations with you.

Connected to Orbit, your AI starts every conversation already knowing who matters, what matters, and what needs your attention. And because the model exists independently of any chat window, your context isn't trapped inside one vendor's assistant: everyone else's memory improves their product, while Orbit's model improves whichever AI you point it at.

## Why it matters: trust, efficiency, and attention

Step back from the machinery and the argument is simple. Your life produces more text than anyone can read, and somewhere in it are the handful of things that actually matter. Every approach to this problem has to answer the same question: who does the reading?

Today, you do. Assembling the state of anything — a project, a trip, a school year — means detective work across emails, texts, and calendar invites: finding the fragments, re-reading them, holding the pieces in your head long enough to make sense of them. A retrieval assistant doesn't change that; it takes in the same messages all over again for every question, paying the full cost of understanding each time. Orbit pays that cost once, when the message arrives — and every question, every surface, and every AI reuses the result.

Handing off that work only helps if you can act on what comes back without checking it yourself — an answer you still have to verify saves you nothing. That's what the living model buys. Orbit's answers aren't fresh guesses over search results; they come from a model that is continuously reconciled against the messages themselves — where a cancellation retracts, a finished thing stays finished, and your corrections are final. When the brief says where things stand, that's where things stand. You act on it and move on.

And that, in the end, is the point: your attention. More arrives every day than anyone could actually read, so it gets skimmed, triaged, and interrupted into your day — and attention researcher Gloria Mark has found each interruption costs roughly 23 minutes of focus. Orbit absorbs the thousands of messages that don't need you and assembles the chaos into the topics that matter — so the scarcest resource you have goes to what matters, who matters, and what needs you.

*—--------*

*Sign up to try Orbit at [orbit.me](https://orbit.me)*