A spatial creative language for the agentic era.

Most 3D engines were created for human programmers writing dense imperative code, or for artists working through visual editors that store large amounts of hidden state. AI agents can use those systems, but they spend too much reasoning budget on plumbing. They often produce verbose code, miss the user's spatial intent, and leave behind projects that are difficult for the next human or agent to maintain.

wave3d.ai starts from a different premise: the authoring language itself should be optimized for humans and agents working together. Direction, time, movement, animation, material, physics, interaction, sensing, state, and placement should be expressed in terms that resemble the way creators naturally describe a scene. The code should read like a compact spatial specification, not a mechanical translation of renderer internals.

That is why WaveEngine favors intent-driven APIs and natural-language builder patterns, backed by serializable one-shot calls under the hood. The goal is not only to make code shorter. It is to make the right code easier for an agent to guess, easier for a human to review, and easier for either one to modify later without re-learning the entire project.

This matters across the places where spatial software is becoming normal: 3D games, digital twins, robotics simulation, XR, and AR. In each case, creators need a way to talk about objects, behaviors, surfaces, controls, and runtime state without drowning in low-level scene graph work. Agents need that even more. If the user asks for a character to move forward, a platform to oscillate, a material to animate, or a scene to use a realistic sky, the code should preserve that intention directly.

Wave Studio is the creative environment built around this idea. It is a GUI-less, code-only, agentic-workflow-driven 3D engine for the new era of spatial computing. Instead of dragging sliders through a traditional visual editor, creators work through code, language, hot reload, runtime inspection, screenshots, and AI agents that can read the project, make precise edits, verify the result, and keep the scene coherent.

The product vision is a workflow where a game level, robot simulation, digital twin, or XR prototype can be authored as a living spatial program. The program is still real code, so it can use modern TypeScript where appropriate. But the center of gravity is no longer generic low-level JavaScript. It is a shared spatial language designed so that human intent, agent reasoning, and runtime performance reinforce each other.