MOTIONVECTOR / THE POSSIBILITIES

A good idea should have a chance to exist.

People have always had more ideas than time, money, or skills. A game that doesn't exist yet. A film you can see in your mind. An experiment you can't afford to run.

We're building a computer that helps people cross that distance—without handing away control of what they make.

Explore possibilities ↘Play The Crossing ↗

The Crossing browser game. A player stands by a ravine where a wooden bridge has been created. — **The Crossing**PLAYABLE BROWSER STUDY

02 / FILM

Direct a scene

03 / SOFTWARE

world.inspect('app')

propose({
  action: 'change',
  target: 'checkout'
})

verify(tests)
commit(approved)

✓ constraints checked
✓ tests recorded
✓ state version updated

Build with agents

01 / HUMANYou decide what matters.

Set the direction, make judgment calls, keep your taste and your say.

02 / INTELLIGENCEAgents help carry the work.

Explore options, perform bounded tasks, and bring back results.

03 / COMPUTINGThe system keeps the record.

Known state, explicit permissions, and evidence of consequential changes.

01 / START WITH A WORLD

Build a place that can change its own story.

Games make the idea tangible. A world has objects, rules, history and consequences. People explore it visually. Agents can work with the underlying state. Both should be subject to the same rules.

GAMES / THE CROSSINGPLAYABLE CONCEPT

A bridge, built with permission.

An explorer reaches a ravine. You ask an agent to build a bridge. The world checks whether that action is allowed, whether the scene has changed, and what the new state will be.

Once the operation is accepted, the bridge becomes part of the world. You can inspect the change and undo it.

Play the world ↗A browser demonstration of these rules.

Illustration from The Crossing: a wooden bridge connects two banks in a stylized mountain game world.

WORLD: THE CROSSING
effect: create bridgeSTATE / VERSIONED
RECEIPT / INSPECTABLE

02 / MAKE THINGS

More of your ideas, out in the world.

A creator shouldn't have to become an expert in every piece of software between an idea and the finished work.

BEAT_04 / SCENE TIMELINE

FILM / MOTION / MUSIC

Direct the film you see in your mind.

Describe an opening shot. Change the pacing. Ask for three versions of a scene. Keep the edit you love, reverse the one you don't, and know which frames were affected.

EVERY EDIT HAS A HISTORYCONCEPT

checkout.spec.ts

describe('checkout', () => {
  it('keeps the price', () => {
    expect(total).toEqual(42)
  })

  it('rejects unapproved spend',
    () => expect(approval).toBe(false))
})

8 / 8 CHECKS PASSED

PROPOSE → VERIFY → COMMIT

SOFTWARE / DEVELOPMENT

Make software without managing every keystroke.

Ask agents to design, implement, test and release a feature. See which files changed, which checks ran, what was approved, and what still needs attention.

ONE CHANGE / THE TESTS BEHIND ITCONCEPT

03 / UNDERSTAND THINGS

Questions can become places to explore.

Some answers deserve more than text. A simulation, experiment or model can let someone test what happens when an assumption changes.

ILLUSTRATIVE MODEL / NOT EXPERIMENTAL DATA

SCIENCE / DISCOVERY

Run the experiment you couldn't run alone.

A researcher asks a question. Agents help explore candidate explanations, run bounded simulations, compare outcomes and preserve the assumptions behind each result.

EACH RESULT KEEPS ITS ASSUMPTIONSCONCEPT

PARAMETER / ORBITAL SPEED

LEARNING / SIMULATIONS

Learn by changing the world.

A student asks why an orbit holds. Instead of memorizing a diagram, they change the starting speed and watch the trajectory respond. They can ask questions of the simulation itself.

LEARNING THROUGH CONSEQUENCESCONCEPT

04 / OPERATE THINGS

Keep the big picture. Delegate the machinery.

People should be able to oversee increasingly complex work without surrendering permission over their resources, money or physical surroundings.

ORDERS / INVENTORY / APPROVALS

ORGANIZATIONS / OPERATIONS

Run something bigger than your team.

A small company can coordinate research, inventory, suppliers and routine operations. An agent may prepare a purchase. The owner still decides whether the money leaves the account.

OWNERS KEEP THE FINAL SAYCONCEPT

PHYSICAL ACTION / BOUNDED ZONE

DEVICES / ROBOTICS

Give intelligence a body with boundaries.

A device can observe the environment, propose a movement and act within explicitly permitted limits. Physical systems raise the stakes: sensing, emergency stops and independent safety systems remain essential.

BOUNDED ACTION / REAL-WORLD SAFETYCONCEPT

THE HUMAN REASON

More people should get to make things that matter to them.

A teacher making a lesson that feels like a discovery. A musician building an instrument that didn't exist. Someone who cannot use a conventional interface, but still wants to work in a powerful creative tool. A small team pursuing a question normally reserved for a much larger lab. We want to expand what people can attempt while preserving their judgment and control.

IMAGINATION × CAPABILITY

05 / ONE IDEA, DIFFERENT WORLDS

Choose a world. See the work beneath it.

The domains differ. The basic questions keep returning: What exists? What is allowed? What changed? What evidence should we trust?

GAMES / A WORLD WITH RULES

Build a bridge. Keep the rules.

An explorer needs to cross a ravine. An agent can propose the bridge while the world decides whether it can be built.

What you ask

“Make the far side reachable.”

What the computer must know

The gap, the terrain, which structures exist and which actor holds build permission.

What you can inspect afterward

A versioned bridge operation, permission decision and resulting world state.

SEMANTIC OPERATION / ILLUSTRATIVE

This explorer illustrates the proposed architecture. It does not execute a model or connect to the production runtime. For a hands-on example of these principles, play The Crossing →

STATE

Know what exists.

Stable identities, relationships and versioned state across supported domains.

ACTION

Express the change.

Operations with explicit targets, inputs and expected effects.

AUTHORITY

Respect the boundary.

An agent proposes work under defined permissions and limits.

EVIDENCE

Show what happened.

Execution records, validation results and recovery where supported.

06 / BUILDING TOWARD THIS

A future needs a foundation.

The examples on this page span working ideas, active engineering and longer-range research. We are building toward them through increasingly capable, testable worlds.

THE FOUNDATION

Media and structured documents.

MotionVector's documented engine work includes a typed document model, deterministic lowering, addressed patches, validation and rendering. Games has a playable browser concept.

THE EXPANSION

Agent-operated software and worlds.

Extend these semantics across game, browser, terminal and other domain runtimes. Make action, permission, state and evidence work across increasingly complex tasks.

THE HORIZON

New forms of computing.

Human–agent collaboration across scientific, organizational and physical environments—each with its own safety, correctness and integration requirements.

MOTIONVECTOR / THE POSSIBILITIES

Let's see what more people can make.

More people should get the chance to try. Here is how we're building the computer underneath those possibilities.

Read the Manifesto ↗Play The Crossing ↗Architecture Note ↗