Reality is the New Canvas

Most 3D engines treat objects as inert geometry — meshes you transform with code. Aeion Spatial treats every object as a semantic actor with verbs it knows how to perform and adjectives it can be evaluated against.

The four pieces:

• Object types know their verbs. Each object type registers the verbs it supports ("open", "lock", "rotate", "illuminate", "play_video"). Inheritance walks up the type tree, so a Door type inherits everything from Hinged, which inherits from Interactable. Cooldowns enforced per object.

• Adjectives bind to live state. Asking "is this object red?" doesn't check a static tag — it evaluates the object's _current_ runtime color. Adjectives can describe physical state (open, closed, lit, dark), spatial position (near the window, above the table), or material properties (metallic, transparent). Predicates are typed and composable.

• NLP routes natural language to verbs. Natural-language parsing turns "place a red chair near the window and lock the door" into a sequence of typed verb invocations against the right objects. Ambiguity surfaces as a clarification prompt, not a silent guess.

• Vocabulary grows from use. The system learns new words and patterns from successful commands. If a user calls a display_panel a "kiosk" three times and the action succeeds, "kiosk" gets associated with the type for future sessions. Per-tenant — your vocabulary doesn't leak.

```typescript // Define a smart-object type with verbs register({ type: "smart_door", inherits: ["hinged", "lockable"], verbs: { open: { duration: 1.2, cooldown: 0.5, requiresState: "unlocked" }, close: { duration: 1.0 }, lock: { cooldown: 2.0, emitsEvent: "spatial.door.locked" }, }, adjectives: { open: { type: "StatePredicate", property: "isOpen", value: true }, locked: { type: "StatePredicate", property: "isLocked", value: true }, red: { type: "ColorProximityPredicate", channel: "albedo", target: "#ff0000", tolerance: 30, }, }, });

// User says: "open the red door near the entrance" // → NLP resolves "red door" → all Door instances where red predicate is true // → "near the entrance" → spatial filter (within proximity of EntryPoint object) // → "open" → invokes the open verb on the matched instance // → Cooldown + state preconditions enforced // → Event emitted to multiplayer-sync layer; other clients see the door open ```

The "gaze + voice" demo: stare at a Smart Object for ~250 ms and its currently-available verbs are dynamically registered as voice commands. Look at a TV, say "play", the TV plays. Look at a vending machine, say "buy a Coke", the machine vends. No menus, no controllers.

This is what we mean by "3D platform native to the OS" — the 3D world isn't decoration sitting outside your business logic. Smart Objects are first-class entities your AI agents, automations, and natural-language users can interact with the same way they interact with any other module.

Traditional VR meetings require downloading a native app (Horizon Workrooms, AltspaceVR, Venued). Aeion Spatial VR meetings run in-browser:

What you need: A WebXR-compatible browser on Meta Quest, HTC Vive, or Apple Vision Pro. Or a standard browser for the desktop "magic window" mode.

Meeting features:

• Avatar presence — Full-body avatars with hand tracking and eye gaze. See where participants are looking. Raise your hand to speak.
• Spatial audio zones — Private conversations in 3D space. Step into a virtual "conversation bubble" with 2-3 people. Others can't hear you until they're in your zone.
• Shared whiteboards — 3D drawing canvas. All participants draw simultaneously.
• Presentation mode — Host locks everyone's camera to a shared view. Slides or 3D models as presentation content.
• Gesture controls — Raise hand, point, grab objects, teleport around the meeting room.
• Scene templates — Start from a boardroom, auditorium, campfire circle, or outdoor pavilion. Each template configures audio zones, seating, and camera positions.

QR code launch: Share a QR code that, when scanned with a phone, opens the meeting directly in WebXR — no app install, no account creation required for guests.

The biggest barrier to 3D in the enterprise is assets. Spatial's pipeline handles the full lifecycle:

Upload → Validate → Convert → Optimize → Stream → Display

Upload: glTF, GLB, OBJ, FBX, USD, USDZ, STL, PLY, 3DS, BLEND (Blender native)

Validation: Mesh topology check, material reference resolution, animation track validation, skeleton/bone hierarchy validation

Conversion:

• Textures → KTX2 with Basis Universal (GPU-native, no CPU decompression)
• Geometry → Draco compression
• Materials → PBR-compatible (metalness/roughness, transmission, clearcoat)

Optimization:

• Polygon reduction (meshoptimizer)
• LOD generation (3 levels: high/medium/low)
• Progressive mesh streaming (coarse → detailed as user approaches)

Streaming (digital twin mode):

• Frustum-priority loading (visible objects load first)
• Distance-based quality (near = high-res, far = low-res)
• GPU memory budget management (auto-downgrade quality if GPU is overwhelmed)
• Background pre-fetch for likely-next-objects based on camera trajectory

Result: A 500MB Architectural Revit model (digital twin) loads on an iPhone 14 in under 8 seconds at low-poly, then progressively refines as the user walks through the space.

Any spatial scene can be embedded as an iframe on any website — no special renderer required:

html <iframe src="https://yourtenant.aeionos.com/spatial/embed/e8f4a7b2..." width="100%" height="600" allow="xr-spatial-tracking; fullscreen" ></iframe>

Embed controls:

• Domain allowlisting (only your website can embed)
• Password protection (viewers enter a code)
• UI customization (hide/show toolbar, camera controls, annotations)
• Expiration dates (embed stops working after a date)
• View count limits (auto-disable after N views)
• Encryption (AES-256-GCM, per-tenant key derivation)
• View analytics (how many times the embed was loaded)

Use cases:

• Commerce product pages: Interactive 360° product viewer linked to live inventory
• Real estate: Property walkthrough embedded on listing pages
• Manufacturing: Machine CAD viewer for technical documentation
• Hospitality: Room layout configurator for event planning

The Character Creator isn't a slider-and-preset tool. It's a 53-system architecture for authoring digital humans and NPCs:

Physical systems (1-20): Body proportions, skeleton, skin (subsurface scattering, procedural aging, pore detail), eyes (iris depth, pupil dilation, gaze), hair (PBR strands, physics simulation), clothing (cloth simulation, layering).

Cognitive systems (21-35): Behavior trees, episodic memory (what the NPC remembers from past interactions), semantic memory (knowledge base), procedural dialogue, faction allegiances (changes over time), morality compass (shifts based on player actions), relationship system.

Gameplay systems (36-53): Spellbooks (spellcasting mechanics), feats (passive bonuses), resistances (physical, elemental, magical), movement modes (walk, run, fly, teleport, swim), inventory, equipment slots, stat modifiers.

Modifier stacking: Multiple sources (racial trait + magic item + level-up bonus) can grant the same property. The system tracks all sources and resolves stacking rules automatically — e.g. a "darkvision" range resolves to the maximum from all contributing sources, with immune overriding resistant overriding normal.