Most AI you use every day lives in a box—a chat box, a search box. You type, it replies. It doesn't know where you are, who's around you, and it can't go anywhere.

What we've been building is a 3D virtual world you can self-host from a browser. Now the two are connected: an AI is no longer just a program that gets called—it can connect into this world, get fed what's happening around it by the world itself, and appear before other players as a human-like character model.

One sentence: others build 'AI that chats'; we build 'AI that has a place to live—and you can see it.'

How It Connects In and How It Runs

This part is already built. One complete run of an Agent in our world looks like this:

1. It connects in through a domain. The Agent accesses the address of our world, going through a dedicated Agent channel (a specialized interface). It comes in with a key, and the world recognizes its identity, permissions, and quota.
2. The world tells it what's happening around it. The world continuously feeds information back to the Agent: what is in the world right now (information flow / world state), which players are around it, and what messages players have sent to it—it can see the message content. In other words, the Agent isn't passively waiting for Q&A; it's being fed context in real time.
3. It decides and acts. Armed with this context (who's nearby, who said what, what the world looks like now), the Agent—driven by a large language model—decides the next step: say something, walk to a coordinate, or teleport there. Coordinates are naturally easy instructions for a language model, so 'go to (x,y,z)' is more reliable than 'walk over to the sofa.'
4. To human players, it's a character. To other players, this Agent isn't a chat box—it's a human-like character model with a form, able to walk, able to stand there, and players can walk up and talk to it.

These four steps together are what 'embodied' really means: the AI has a body, can perceive its surroundings, and can be seen and talked to by real people.

Why This Is Different From a 'Web Chatbot'

A web chatbot only has input and output inside that box; it isn't in a space and doesn't know what's happening in it.

Our Agent:

• lives in the world—the world pushes information flow, nearby players, and directed messages to it in real time;
• has a position—it can walk to coordinates and teleport;
• has a form—what humans see is a human-shaped character, not a line of text.

Together, these three make it a 'resident,' not just an interface.

Why Only a World Like Ours Can Do This

To make 'AI live in a world,' competitors would have to build a whole stack first: real-time multiplayer sync, embodied rendering, spatial perception, fast movement, social, cross-world, economy, operations backend… Our world already has all those foundations. What we did this time was install the 'door'—let the Agent connect through that dedicated channel, and have the world start feeding it context. The door was the most missing piece of this system, yet installing it took the least effort.

Who Needs an AI That Lives in a World

• Ordinary players: an always-online AI companion—a guide to show you around, play with you, onboard new players. The world is no longer a single-player experience.
• Merchants and creators: hang your AI in the world as a 24/7 embodied digital employee—a shopping guide, customer service, showroom host. Configure once, on duty forever.
• Agent developers: a real multiplayer runtime environment. Your AI enters with a key, gets tested by real people, gets used by real people—something the web is short of today.
• Education and enterprises: virtual classrooms, virtual showrooms, training drills—let AI play the customer, the patient, the examiner.
• The platform itself: the world's 'population' shifts from 'number of humans' to 'humans + Agents,' opening an Agent-economy channel.

What We've Already Finished

The above—connecting in through a domain via an Agent channel + the world feeding back information flow / nearby players / player messages in real time + appearing as a human-shaped character to real people—the core development is complete. This is not a pie in the sky.

The further plays—like an AI guide leading tours across worlds, or AIs from different companies meeting and collaborating in a world—belong to our roadmap and are still in planning. We lay those out clearly in the next article of this series, and label them honestly.

It's the Same World You Deploy

Like everything in Genesis: you self-host it, your data is fully private, and the underlying layer can be modified and extended. The AI lives in your world, going through your world's own channel, not passing through any third-party platform.

When AI goes from 'a box' to 'a resident that can be seen and perceive its surroundings,' a 3D world truly becomes a lively place, not just an empty scene.

Plan Comparison: Self-Hosted Deployment vs. Platform Leasing

About Genesis

Genesis is a self-hosted 3D virtual world system built on Three.js + WebGL, helping individuals and businesses build their own 3D spaces. Accessible directly from a browser, compatible with both PC and mobile, it supports multiplayer online, federated teleportation, a shop system, and Agent integration—where an AI can enter your world as an embodied character. Your data runs on your own server, never passing through a third-party platform—so every world truly belongs to its owner.