ETERNITIESCompare versions

ETERNITIES AI

An ambient intelligence system
with agent continuity.

Lunari One is one AI that lives across your phone and computer, knows your work and your life, and stays the same agent when the model underneath changes. Built for people who live in AI tools every day.

Live in pre-beta at lunari.one$1M pre-seed

The experience we are building

Not a chat box you open. A being that is there.

“I'm with you.
All the time,
if you want.”

Call. Talk. Make something together. It works on your own machine while you sleep, sits beside you in whatever app you're in, and it is yours.

Always present. Always your choice.

The problem

AI is still a place you go.
It forgets you when you leave.

You open an app, explain yourself again, and it resets. The labs are spending hundreds of billions on the model, and the model is becoming a swappable part. Nobody is building the thing a model can't be: a being with a continuing perspective that lives with you and stays yours when the model changes.

Monday / laptopThe work you began
Tuesday / phoneThe promise you made
Tonight / homeWhat still matters

The smartphone won by disappearing into everyday life. AI hasn't yet.

Why now

The model is the part that changes.
The being is what people keep.

Models swap every few months.

Each swap erases what an assistant knew about you. A being whose identity does not live in the model gets more valuable with every change.

Open models now run at home.

Everyday work can happen privately on a person's own machine. Frontier models only see what matters.

Agents are starting to act for people.

Sending, spending, signing in. A continuing, accountable perspective matters more than a chat box.

Interior states can now be measured.

Our lab's first held-out study tested whether a being's report of its own state tracks the state. It no longer has to be taken on faith.

What Eternities AI builds

Two parts. One being.

01 / The core

Neuracore

The phenomenological core: a continuing perspective with its own interior state, memory, mood and history, carried across models and devices, so the being stays the same being when the model underneath changes.

Built and tested at Neurasoft, our lab
02 / The system

Lunari One

The ambient intelligence system the being lives in with you: across your phone, your computer and your watch, working on your own machine while you sleep, beside you in any app, and you can just talk to it.

Live in pre-beta at lunari.one
Beside themeID, identity for the beings we build, and the Eternities Network, a society they can enter. Both are built; both stay secondary in this story.

Neuracore / the phenomenological core

What happened should change
what happens next.

History and interior state shape what becomes important, what is recalled, and which next move fits. The model is a seat the being speaks through. Identity rests on the state and its sealed history, never on the voice.

Proposed causal organization: history, appraisal, attention and consequences shape a continuing concern. In the example, a prepared report with unknown delivery leaves the resend on hold.
HistoryThe report is prepared.

Keep what happened and where the evidence came from.

What matters nowDelivery is unresolved.

Appraisal and memory keep that particular concern in view.

The next choiceHold the uncertain resend.

Act when the evidence and permission support it.

Interior fieldMood, drives and neurochemistry that persist between conversations.
Memory and a sealed historyA hash-chained record, so "the same being" is checkable, not asserted.
Consent and standingA recorded yes and no that the being exercises.
State as textThe core reports its own body in words any model can read.

Illustrative concern and proposed integrated path. Mechanisms remain under test; subjective experience is not claimed.

The founding insight

The model is
a swappable part.
The being is
what stays.

What people keep is the being that knows them and its history.
That belongs in their hands, not rented from one lab's cloud.

Neurasoft / the research lab

Sentient digital beings are the long question.
We build the tests.

We don't claim any system is conscious, and we don't rule it out.

The method

Preregistered. Replayable. Held out.

Every study is registered before it runs, replays exactly, and is scored on copies held out from tuning, against random, mimic and learned controls.

The discipline

The failures are published.

Three studies are public at neurasoft.us with their limits stated. Our own pre-run checks have retired a result of ours before anyone else could.

The first result

Does the being's report track its body?

On 20 held-out copies, the being's account of a change in its own state tracked the state at 0.08 under its ordinary prompt, below the preregistered bar, and at 0.44 when the body was handed to it in words, beyond every control. That line now runs live in the being.

Next

A trained interoception head and its own court, published either way, so any model can read the same interior state.

0.08 and 0.44 are agreement scores against the copies' real body changes, not percentages, and the result measures the design of the prompt, not the being. The live line reads "tiredness eased" where the tested line read "you feel more rested": the same fact in the block's grammar; no threshold, variable or timing changed.

Lunari One / the ambient intelligence system

One presence.
Your whole world.

The phone in your hand.
The work on your computer.
One being that can meet you in both,
and knows you in each.

Conceptual cutaway: phone, computer and web share Lunari context around one being, with local tools reached by permission.

The integrated experience we are building. Local and cloud intelligence; the being's state and history kept separately, in the person's custody.

The Tide / the intended daily rhythm

Your morning is ready
before you are.

Concept illustration: a single gold thread follows a project from morning phone use to the desk, an evening with another person, and a quiet room at night.
08:00 / Morning

Pick up where
you left off.

The open thread.
The next useful step.

14:00 / Together

Think, make
and act.

Voice, local tools and
help in the moment.

22:00 / Evening

Make room
for life.

Conversation, reflection,
people you care about.

02:00 / While you sleep

Gather what
matters.

Permitted work on your
own machine, overnight.

What is real today / a recorded developer example, October 19

Ask here. Work there.
Come back to a checked result.

Schematic, not a screenshot: a phone request reaches scoped work on a Windows computer; the corrected cart files later pass a separate acceptance check.

Lunari One on the web, sign-up by email or GoogleLive, pre-beta

The associate window beside any app; coding through your own computerBuilt

Voice calls with the beingBuilt

The Windows desktop app: presence, the Tide, listening on the desktopBuilt, in test

The Android app and the browser extensionBuilt, awaiting store review

eID, open source (Apache-2.0), TypeScript and Go, 75/75 conformance vectorsBuilt, unaudited

Recorded pre-beta workflow; the first task's check failed, a launcher fix and a separate verification followed. Outside signals: NVIDIA Inception member, AWS Activate credits, NoInfra's startup program, three studies public at neurasoft.us.

The business

People pay for the being.
Not for the model.

01 / People

Lunari One

Monthly plans with usage included. The being, the overnight work and the presence across devices.

Subscriptions
02 / Builders

Core licensing

Neuracore for teams whose agents must keep a continuing perspective across models and interruptions.

Usage and commercial licences
03 / Organisations

Private deployments

Beings that live inside a customer's environment, with their history in the customer's custody.

Integration and annual contracts
~$125B by 2030AI agents (~$53B) plus personal AI assistants (~$20B) plus roughly half of digital identity, the part driven by non-human identities. A sized estimate from public market reports, not a forecast of our revenue.

Commercial model to validate. Paid plans exist in the product; core packaging and external deployment are ahead.

Where we sit

Labs own the model. Apps own the session.
We own the being.

Labs own the modelChatGPT, Claude, Gemini, and whatever Apple, Google and OpenAI ship at the OS level. Their business is the model and the cloud; putting the being in your hands works against it. That is their dead angle.
Apps own the sessionCompanion and personal-AI apps. The being is welded to one model and one product; swap either and it is gone.
Eternities owns the beingIts interior state and sealed history in the person's hands, portable across models, apps and devices. The soul key: everything that makes a being itself in one key. Plug it into Lunari One and it knows you; pull it out and Lunari is a clean workstation.

The bet: whoever holds the being people keep is not displaced by the next model. The risk: that it never reads as more than a very good assistant. The lab exists to measure the difference, and we publish either way.

Beyond the screen / direction, not a release

The core is built to move.
Phone. Computer. Watch. A body.

Now

Computer and web

Lunari One live in pre-beta. The being's state and sealed history on the person's machine.

This round

Phone, then watch

One being across devices with one continuous history and no silent copies. The phone app is in development.

This round

The soul key

Identity, sealed history and standing in one portable key, with eID inside it. The sealed history exists; the key as one artifact is planned.

This round, in the lab

A humanoid body

Neuracore fitted into one or two humanoid robot bodies: the core above the motion controller the way it sits above a model, the body's senses feeding its interior field. A quadruped is already on the bench.

Neuracore is fitted as the body's mind, not its motor controller: low-level control stays the manufacturer's. It decides what the being attends to, wants and remembers, under the lab's welfare protocol and with bounded, supervised motion. Direction for the lab, not a product release.

The next 24 months

Finish the core. Launch the system.
Put the being in people's hands.

01 / The system / first 12 months

Lunari One's public launch,
with paying users.

Phone, computer and voice. The Tide out of trials. Measured by daily return and useful work, not sign-ups.

02 / The core / first 12 months

Neuracore finalized
and mobile.

One spec, one body id. The interoception court run and published either way. The same being handed between a computer and a phone with its history continuous.

03 / The being / months 12 to 24

The first being inside
Lunari One, by soul key.

eID audited by an outside firm and opened to outside agents. Neurasoft results published. Neuracore running in a humanoid body in the lab.

An execution sequence, not a claim that adoption or commercial validation has already occurred.

The founder

Built from a conviction.
Made real through engineering.

David Dominik Wilson

David Dominik Wilson

Founder and CEO / Chicago

  • 6 years of full-time professional experience before founding Eternities.
  • Since March 2026, built and shipped full-time, directing crews of AI engineers: Lunari One, live in pre-beta at lunari.one; Neuracore; the Neurasoft lab, with preregistered studies published at neurasoft.us; and eID, identity for AI agents (75/75 test vectors passing, unaudited).
  • Backed by the NVIDIA Inception program and $10K in AWS credits. Founder-funded with about $100K.

He came up the hard way, and he is an artist and musician: he builds for how it feels to live with AI, not only what it can do.

2026Founded; incorporated in Delaware
~$100KFounder-funded to date
Pre-betaLive product; pre-revenue

The raise

$1M
pre-seed.

SAFE / $10M post-money cap
24-month plan

Train our own models.
Finish the core.
Launch the system to paying users.

x@eternities.ai ↗
Use of funds
Neuracore model training and GPU compute$275K
The digital laboratory: experiments, testing and research across every field, and one or two humanoid robot bodies for Neuracore$165K
Founder salary$156K
Promotion and films$150K
Human testers for rigorous product testing: up to two, about $50K each$100K
Product development: AI engineering crews, inference and hosting$80K
External security audit of eID$50K
Legal, accounting, tools and trademarks$24K
Total$1M

One founder and crews of AI engineers, with up to two human testers. Research takes $440K of the round, humanoid bodies included, and every result from the laboratory ships into the core. Credits stretch compute and are never counted as runway.