Greetings, users.
For years Virtual Reality (VR) had a fairly obvious problem: building anything interesting required the patience of a medieval cathedral mason, except the cathedral was made of polygons and every third software update broke the door.
Artificial Intelligence (AI) is changing that remarkably quickly.
The important convergence between VR and AI isn’t really about putting ChatGPT inside a headset. It is AI becoming the creation layer of spatial computing: generating environments, objects, code, characters, voices and eventually whole interactive worlds.
Meta is already there. Its Horizon creator stack can generate meshes, textures, sound effects, skyboxes, scripts and environments, while AI-powered Creator Assistant helps assemble experiences. In September 2026 Meta went further, announcing Horizon Create and Horizon Studio, where creators describe a 2D or 3D game in natural language and AI constructs the underlying experience.
Even VR development itself is becoming agentic. Meta’s latest developer tools allow coding agents to install builds, interrogate documentation, inspect performance and interact with VR devices. In other words, we’re moving from “learn this enormous Software Development Kit” to “tell the machine what you’re attempting and let it help operate the machinery.”
Then there is Google. Android XR—the Extended Reality operating system developed with Samsung and Qualcomm—puts Gemini directly inside spatial computing. On Samsung’s Galaxy XR, AI can understand more of what the wearer is doing and provide contextual assistance while they work, explore or play. Google is carrying the same idea into smart glasses.
This matters because AI suddenly has space to inhabit. Your assistant stops being a box into which you type queries and becomes a persistent actor that can see, hear and respond inside an environment.
The Non-Player Character (NPC) is about to get an upgrade.
NVIDIA’s Avatar Cloud Engine (ACE), for example, combines generative models, speech systems and animation technology to create digital humans capable of real-time interaction. The obvious destination is a world populated not by NPCs endlessly informing you that somebody stole their sweetroll, but by characters capable of conversation, memory and semi-autonomous behaviour.
Meanwhile Roblox is attacking the other bottleneck: making 3D things. Its Cube foundation model can generate 3D assets, and in 2026 Roblox introduced “4D generation”: objects with behaviour attached. Not simply make me a car, but make me a car whose wheels turn and which can actually be driven. Roblox says its eventual target is complete scenes containing environments, code and animation generated from natural language. Significantly, it has also open-sourced Cube 3D for use outside Roblox.
So yes, the barrier to creating a virtual world is collapsing.
Which raises the traditional MetaRider question:
Who owns the bloody thing?
Convenience has always been the gateway drug of platform dependence. Build your identity, audience, commerce and assets entirely inside somebody else’s world and you haven’t created a digital nation. You’ve opened a concession stand in their airport.
That doesn’t mean “Meta steals your IP.” Its current developer agreement says each party retains its intellectual-property rights, although developers grant Meta the licences necessary to distribute and market their products. The bigger issue is structural: discovery, payments, rules and customer relationships become dependent upon the platform.
VRChat illustrates the economics neatly. Its paid-subscription model has described an approximate split in which 50% reaches creators, 30% goes to the underlying platform and 20% goes to VRChat and partners. Nobody is behaving mysteriously here. Platforms provide users, infrastructure and payments; they take their tithe.
For somebody entering VR today, however, there is another route.
Start with Blender + Godot + OpenXR, and use WebXR where browser distribution makes sense.
Blender states explicitly that the artwork you create belongs to you. Godot is open-source under the MIT licence, charges no royalties and says the games users create are theirs. OpenXR is a royalty-free open standard designed to reduce dependence on a single headset vendor, while WebXR brings immersive experiences onto the open web.
You can still buy a mainstream headset. You can still publish to major stores. You can even use platform AI tools when they make sense.
Just keep your source code, assets, identity and audience portable.
Treat the headset as a terminal, not a landlord.
Because the most interesting possibility created by AI isn’t Meta building the metaverse faster.
It’s you being able to build one without asking Meta—or anyone else—for permission.
Ride or die ??