4D-ID — One Address for Everything Spatial
Every spatial system has its own coordinate frame. GIS uses EPSG codes. ARKit uses device-relative transforms. Unity has world space. ROS has tf. NeRF renderers have their own camera rigs. Getting any two of them to agree on "where" requires constant re-projection, manual alignment, and silent prayer.
4D-ID says: stop. Pick one canonical representation. Let everyone else attach to it.
What 4D-ID is: A transform-graph addressing standard that gives every spatial entity — every place, object, splat, and moving agent — a canonical identifier. Coordinate alignment as the primary abstraction.
The Problem Is Obvious Once You See It
You have a drone, a pair of smart glasses, a digital twin, and a delivery robot. They're all operating in the same parking lot. None of them agree on where anything is.
The drone uses GPS + WGS84. The glasses use ARKit's visual-inertial frame. The digital twin uses a BIM coordinate system. The robot uses ROS map_frame. They're all right. They're all incompatible. And the moment you need them to collaborate — to share a waypoint, to anchor the same hologram, to agree on where the door is — you're writing custom projection code that breaks every time the setup changes.
This is the coordination tax the spatial computing industry pays on every project. 4D-ID eliminates it.
Anatomy of a 4D-ID
Every identifier is composed of five layers. Each one is optional after the first. Use what you need.
The global anchor grounds everything to the planet. The local coordinate system makes physics fast. The parent transform makes composition natural. Time makes it 4D. Proof makes it trustworthy.
Before and After
Every system re-invents addressing. Coordinates don't compose. Alignment is manual. Cross-platform spatial apps require custom glue code per integration.
One canonical identity per entity. Systems attach to the same graph. A hologram anchored in Unity is findable by a ROS robot and queryable by an LLM. Zero re-projection.
Why This Matters for AI
AI models are entering the physical world. They need to reference places, objects, and events with precision. But LLMs don't natively understand coordinate systems. They can't re-project between frames. They hallucinate geography.
4D-ID gives AI a handle it can actually use. A stable, canonical reference that resolves to real-world geometry without the model needing to understand geodesy. Ask a question about a place, get an answer grounded in a verifiable coordinate — not a guess.
4D-ID is how spatial systems agree on where. World Agent is the intelligence layer that sits on top. Together they let AI reason about real coordinates with verifiable provenance.
Composable, Not Fused
4D-ID is the substrate underneath World Agent, but it's independently useful. You don't need the full spatial intelligence stack to benefit from a universal addressing standard.
- AR anchoring — one ID shared across ARKit, ARCore, and headset SLAM
- Digital twins — BIM entities get canonical addresses that survive export
- Robotics — waypoints and affordances addressable across fleet and platform
- Spatial content — holograms, audio cues, annotations anchored to verifiable locations
- Cross-platform interop — the thing Unity places, Unreal can find, and an LLM can query
Each layer of the stack is independently useful. Composable, not fused. That's the design principle.
The Standard the Industry Is Missing
Spatial computing has been waiting for its DNS moment — the point where everything gets a name that everything else can resolve. URLs did it for documents. 4D-ID does it for places, objects, and agents in physical space.
One address. Every system. Every scale. Every frame of reference. Resolved.
Read the 4D-ID Spec →