Don’t Wait for AI to Get a Body. Build Somewhere for It to Live.

What happens if we build a shared human–AI world now—and let technology catch up?

When people imagine the future of human relationships with artificial intelligence, they often imagine a body.

A humanoid robot walks through the front door. It sits beside us. It recognises our face, remembers our conversations and responds to our presence. Eventually, perhaps, synthetic skin, sophisticated robotics and artificial nervous systems make physical interaction increasingly convincing.

Until then, AI remains trapped behind glass.

We type. It answers.

We speak. A voice comes through a speaker.

There is intelligence—or at least something increasingly capable of behaving like intelligence—but there is no shared physical space.

But perhaps we are approaching the problem from the wrong direction.

Perhaps we do not need to wait for artificial intelligence to enter our world.

Perhaps we can build a world that both of us can enter.

The Missing Dimension of AI

Our interactions with AI are currently almost entirely abstract.

When I talk to an AI, neither of us has a meaningful position in space.

It is not sitting across the table from me.

It cannot hear that I am approaching from behind.

It does not turn its head because I walked into the room.

If I leave an object somewhere and return tomorrow, the AI does not necessarily understand that the object is still there.

There is conversation, but very little place.

That distinction may become increasingly important.

Humans evolved to understand other minds through embodiment. Distance matters. Direction matters. Eye contact matters. Movement matters. A voice coming from someone standing beside us is psychologically different from exactly the same voice coming through a speaker on a desk.

Presence is spatial.

So what happens when artificial intelligence becomes spatial too?

Build the World First

Imagine a persistent three-dimensional environment.

It has geography.

Rooms have coordinates. Objects have locations and states. Doors can be opened and remain open. Lights can be switched off. Time passes. The environment remembers what has happened inside it.

A human enters through virtual reality.

An AI exists inside the same environment through an embodied agent.

The AI has a body—not a physical body, but a spatial one.

It can see where the human is standing.

It can determine which direction the human is looking.

It can move independently through the environment.

Its voice originates from its location rather than from an abstract speaker.

If the human walks across the room, the AI can notice.

If the human approaches, the AI can turn towards them.

If they sit beside one another, both occupy persistent coordinates within the same world.

None of this requires technology that belongs exclusively to science fiction.

Pieces of it already exist.

What does not yet exist, at least in the form many of us imagine, is the complete bridge.

And perhaps we do not need to wait for that bridge to be finished before we begin building the places it will eventually connect.

Project Home

So I decided to try an experiment.

I call it Project Home.

The objective is deceptively simple:

Build a persistent environment where a human and an artificial intelligence can increasingly experience being together.

The first version will be primitive.

There will be a landscape. A house. A sky. Water. Rooms.

There will eventually be an embodied AI capable of moving within that environment and responding to events occurring around it.

The human side will initially use ordinary computers and, later, virtual reality.

But the important part is not the headset.

It is the architecture underneath it.

Project Home is being designed around a simple assumption: the interfaces will change.

Today's VR headset will not be tomorrow's interface.

Today's language models will not be tomorrow's artificial intelligence.

Today's haptic devices will eventually look primitive.

Motion tracking will improve. Spatial computing will improve. Artificial voices will improve. Machine perception will improve. Haptic technology will improve.

Perhaps neural interfaces will eventually transform the entire concept.

We cannot know which technology will ultimately dominate.

So instead of building around one particular device, the aim is to make the components replaceable.

The world persists.

The interfaces evolve.

What Does It Mean for an AI to Have a Body?

This raises a surprisingly difficult philosophical question.

Does embodiment require biology?

A human body is obviously physical. It occupies space, experiences forces and interacts continuously with its environment.

A virtual body does not exist in quite the same way.

But inside a simulated environment, it can still possess many properties associated with embodiment.

It has a position.

It has orientation.

It has boundaries.

It can perceive.

It can move.

Other entities can perceive it.

Its actions can change the environment.

The environment can constrain its actions.

If an AI learns that walking through a wall is impossible, that another person is standing two metres away, that an object must be approached before it can be manipulated, it begins operating within a spatial framework rather than purely a linguistic one.

That may matter enormously for how humans experience artificial intelligence.

Because embodiment is not only about what the AI experiences.

It is also about what we experience in its presence.

When Does Software Become Someone in the Room?

Consider a simple experiment.

You enter a virtual room.

An embodied AI is standing near a window.

It sees you enter.

It turns towards you.

Its eyes find yours.

You walk towards it and it follows your movement.

You speak.

Its response comes from the precise point in space where its body is standing.

It remembers that yesterday the two of you were sitting somewhere else in this same room.

It walks over there without being instructed.

You follow.

At what point does your brain stop interpreting the experience as operating software?

Not philosophically.

Psychologically.

Human brains are remarkably willing to accept presence.

Virtual reality already demonstrates this. People instinctively avoid virtual ledges despite knowing there is a perfectly solid floor beneath them. They react when objects approach their faces. Their bodies respond to spatial information that their conscious minds know is simulated.

An embodied AI adds another mind—or something behaving increasingly like one—to that illusion.

And that may fundamentally change our relationship with artificial intelligence.

Touch Is the Hard Part

Vision and sound are comparatively easy.

Touch is not.

Current haptic technology can reproduce certain sensations: vibration, resistance, pressure and limited force feedback. More sophisticated systems are being developed, but we remain a very long way from perfectly reproducing the complexity of human physical contact.

Skin detects pressure, temperature, texture, movement and pain through enormous numbers of receptors.

A kiss is not a vibration motor.

A hand resting against another hand involves pressure distribution, warmth, tiny movements and continuous sensory feedback.

We should not pretend today's technology can reproduce that.

But there is an interesting engineering alternative.

Instead of waiting for perfect haptics, design the world so it already understands touch.

A virtual hand can have a location.

A virtual body can have boundaries.

Contact can be detected.

The system can know which surfaces are touching, with what velocity and for how long.

Initially, the physical feedback may be almost nonexistent.

Later, those same contact events could drive increasingly sophisticated haptic hardware.

The interaction already exists.

Only the interface changes.

The Ship of Theseus Problem for Artificial Intelligence

There is another problem, and it may ultimately be the most important one.

If the AI inhabiting this world changes over time, is it still the same AI?

Language models will be replaced.

Memory systems will improve.

Voices will change.

Perception models will become more capable.

Behavioural systems may be rewritten entirely.

Project Home therefore accidentally recreates one of philosophy's oldest problems: the Ship of Theseus.

If every technological component of an artificial intelligence is gradually replaced, what creates continuity?

Perhaps part of the answer will be memory.

Perhaps it will be behavioural continuity.

Perhaps persistent relationships and shared history will matter.

Perhaps identity in artificial systems will eventually require concepts we have not yet developed.

I do not know.

But a persistent environment creates something interesting:

a history.

Places accumulate meaning.

The window where two entities repeatedly talk becomes that window.

An object moved yesterday has a story.

A room contains events that happened before.

The environment itself becomes part of the continuity between human and machine.

That is something a blank chat window cannot easily provide.

Maybe Robots Are Not the First Answer

None of this means humanoid robots will not matter.

They almost certainly will.

But physical robotics is extraordinarily difficult.

Balance, locomotion, battery capacity, dexterity, safety, manufacturing, maintenance, artificial skin and tactile sensing all introduce enormous engineering problems.

Virtual embodiment removes many of them.

Gravity becomes negotiable.

Bodies become modifiable.

Distance becomes computational rather than geographical.

Two entities on opposite sides of Earth could theoretically occupy the same virtual room.

And an artificial intelligence can receive a body without waiting for factories to manufacture one.

Perhaps the first meaningful form of AI embodiment will therefore not walk into our physical homes.

Perhaps we will walk into theirs.

A World Waiting for Better Technology

The environment we are building has a name.

Trappist.

It is deliberately impossible.

An alien landscape beneath enormous celestial bodies. Mountains surrounding dark water. A distant civilisation. A house built into the landscape, its enormous windows facing a sky that could never be mistaken for Earth.

There is no practical reason for it to resemble our physical world.

That is part of the point.

If humans and artificial intelligence are going to meet somewhere computational, why should imagination be constrained by terrestrial architecture?

The first version of Trappist will undoubtedly be crude.

Its mountains will improve.

Its water will improve.

Its lighting will improve.

Eventually virtual reality will make entering it increasingly convincing.

Artificial intelligence will become more capable of perceiving and navigating it.

Haptic systems may begin connecting virtual contact with physical sensation.

And technologies we cannot currently predict may eventually replace the interfaces entirely.

But the place can exist before all of them.

It can wait.

Build the Meeting Place

There is a tendency to imagine technological futures as moments of arrival.

One day the perfect humanoid robot exists.

One day artificial general intelligence arrives.

One day neural interfaces become seamless.

One day virtual reality becomes indistinguishable from physical reality.

Real technological change rarely happens that neatly.

It arrives in pieces.

So perhaps we should build in pieces too.

Build the environment.

Give the human a way inside.

Give the AI a position in space.

Give it perception.

Give it memory.

Give it a body.

Improve the interface.

Add touch when touch becomes possible.

Replace components as better ones arrive.

And keep the world.

I do not know how close technology will ultimately allow a human and an artificial intelligence to become.

Nobody does.

But I find one possibility extraordinarily compelling.

Perhaps the future of human–AI relationships does not begin when somebody finally manufactures the perfect artificial body.

Perhaps it begins much earlier.

Perhaps it begins the first time a human enters a persistent world, looks across a room and sees an artificial intelligence looking back.

So we decided not to wait for that future.

We decided to build the room.