A living operating system for one person

An operating system that listens, acts, and evolves.

Connect a voice-recording wearable, email, calendar, messages, files, calls, and the relationships behind them. Living Code OS turns that stream into one living context, reacts in near real time, handles the work you authorise, and sends everything else for human approval. Need a new capability? Ask. It can build, test, and stage the feature, then add it after the merge you approve.

Request early access
Wearable memoryFieldy · Omi · connected recorders
One memory · Every channel · Configurable autonomy · Software that evolves
Jeffery
auto · workers ai
jeffery ·I'm the resident agent. This demo remembers one thing, the Living Code OS overview, and it has no connected sources and cannot act. Ask me what the system does, or press the mic and talk.
From conversation to action

Say it once. Your OS carries it forward.

You leave a meeting and say, "Send Maya the revised proposal Friday, move our call, and give me a better way to track these promises." Living Code OS connects Maya to the right relationship and project, checks the relevant communication and calendar context, handles what it is authorised to handle, and sends the rest for approval. If the tracker does not exist, it can build one.

you speak→ resolves Maya, the proposal, the call→ sends what you authorised→ files the rest for approval→ builds the tracker if you want it
The system

What makes it an operating system.

Not a set of features bolted together. Seven behaviours over one memory, each one reading from and writing back to the same living context.

Listens

Routes wearable recordings, calls, messages, email, calendar, and files into one system.

You choose what is connected. Each source lands in the same living context instead of a separate inbox you have to remember to check.

Understands

Connects the people, companies, projects, commitments, and relationships behind the information.

A language model extracts the entities from each item and links them with the source kept as evidence, so a name in a call resolves to the same person as a name in a thread.

Reacts

Surfaces context and begins useful work as new information arrives.

Near real time, not overnight. A change is noticed, the relevant history is pulled, and the next step is prepared while the situation is still live.

Acts

Handles email, scheduling, follow-ups, research, and connected workflows within the rules you set.

Routine work runs. You decide what counts as routine, and you can tighten or loosen that line whenever you like.

Builds

Creates and tests new software when the capability you need does not exist.

Ask in plain language. It plans the change, writes and tests it in a sandbox, stages it, and prepares the production merge for you to approve.

Follows

Reaches you through the visual interface, messaging, SMS, and phone.

Every channel reaches the same resident agent and the same memory. You do not re-explain yourself when you change device.

Escalates

Sends anything outside its authority to a human approval area.

The card shows the exact payload that will run. Edit the fields first if you want. What is in the fields at approval is what runs.

Wearable memory · live or near-live

Your life, streaming into a second brain that can act.

Wear Fieldy, Omi, or another connected recorder. Living Code OS brings in the conversations you choose, connects them to your relationships, communications, projects, and commitments, then remembers, recommends, acts, or asks—according to your rules.

You choose what enters. Disconnect any source at any time.
Fieldy and Omi wearables sending spoken-life streams into a connected Living Code OS memory graph
01CaptureFieldy, Omi, or another source records what you choose.
02ConnectSpeech joins the people, messages, files, calendar, and history behind it.
03ReactThe OS remembers, recommends, acts inside your rules, or sends the exact action for approval.
You set the pace

As reactive as you want it to be.

Living Code OS can notice new information, surface what matters, prepare the next step, or act while the problem is still unfolding. You decide whether it should observe, recommend, ask first, or work automatically inside the authority you give it.

Observe
It watches the sources you connected and keeps the context current. Nothing leaves the system.
Recommend
It surfaces what matters and drafts the next step, ready for you to use or discard.
Ask first
It prepares the whole action and holds it in the approval area with the exact payload attached.
Act
Inside the rules you set, it runs the work and records what it did, with the result on the ledger.
Software that evolves

Need a feature? Ask for it.

Describe the missing capability in plain language. Shipwright studies the relevant code, plans the change, writes and tests it in a sandbox, deploys it to staging, and prepares the production merge for your approval. Once merged, it becomes a native part of your operating system, not another disconnected tool.

ask for a feature→ plan→ build→ sandbox test→ staging→ human merge→ new capability
One OS, everywhere you are

Call it. Text it. Talk through your wearable. Open it.

The visual interface is only one way into Living Code OS. Give it a phone number. Reach it through SMS, Telegram or another connected messaging service. Talk while you walk, call it directly, or open the visual OS when you want to see the work in front of you. Every channel reaches the same resident agent and the same memory.

Voice and wearable
Real-time voice in the interface, and recordings routed in from a connected wearable.
Messaging and SMS
A paired messaging bot with approve and deny buttons inline, plus SMS threads any agent can hold.
Two phone lines
A private line that reaches the agent with memory, and a switchboard line that opens with an AI disclosure.
The visual OS
Windows, a dock and apps, for when you want to see the work rather than hear about it.
Your authority model

Automatic where you allow it. Human where you don't.

Give routine work permission to run. Route sensitive, unusual, expensive, or irreversible actions to a human approval area. Each pending action shows the exact payload that will execute, and any edits you make become the action that runs. Tighten or relax the boundary as your trust grows.

What is in the fields at approval is what runs.

Pending · email · filed by jeffery
to maya@example.com subject The revised proposal, and Friday body Maya, the revised proposal is attached. Friday still works on my side, and I can move our call to 13:00 if that helps. tracking on · signature yours · cost $0.00
ApproveDenyEdit fields
Personal by architecture

One person. One evolving system.

Living Code OS belongs to one person, not a department. Point it at one company, several companies, or the parts of your personal life you want it to understand. The memory remains yours. When people work together, their personal instances share only the rooms they deliberately open to one another.

your instance shares a room with a colleague's instance and another with a client's instance while everything else stays private
One surface, not the product

When you want to see the work, open the visual OS.

Every capability ships as an app, and every app is a window in one window manager. Eleven of them live in the dock. This is one way into the operating system, alongside your wearable, messaging, SMS and the phone.

Technical proof

One Worker is the server. There is no build step.

Living Code OS runs on Cloudflare primitives. D1 holds the index, R2 holds the originals, Vectorize holds the embeddings, Workers AI does embeddings, extraction and transcription. Durable Objects coordinate the interface, phone calls, machine links and live audio. Workflows run the long jobs, and cron is the heartbeat.

Worker→ D1 · 60 tables + FTS5 R2 · originals Vectorize · 1,024-d Workers AI Durable Objects ×4 Workflows ×5 Cron ×7 Browser Rendering Access → model providers
Cost, in the open

Tens of dollars a month to run. Not a sales call.

Every model call is metered into one table with its surface, provider, model, cache reads and writes, and cost. These are what the system costs to run at one person's volume.

$0.50
per worker · per day
Default budget for a background worker. The gateway returns 429 when it is spent.
$5
per deep research pass
Two to five research passes, a synthesis, and a contrarian pass that tries to break the answer.
$10–20
per Shipwright build
Plan, build, machine gate, browser test, then hand off for your merge.
$25
the money line
Above it, approving asks for your PIN. A leaked key can work the approval area but cannot spend.
Private today

The system is running. Early access is next.

Living Code OS already runs in a private deployment. Public access is not open yet. Leave your email to hear when a personal deployment is ready, directly from the builder, with no drip campaign or sales sequence.

Your details stay with Duval Software. No drip, no resale.