Introduction: merging spatial computing and artificial intelligence
What is spatial computing and why does it matter for AI assistants
Spatial computing extends computation beyond the 2D screen by digitally mapping the physical world. Key devices include:
- AR/VR headsets (Meta Quest Pro, Apple Vision Pro, Samsung HoloLens)
- Depth sensors and LiDAR integrated into smartphones
- Spatial maps based on GPS and SLAM (Simultaneous Localization and Mapping)
For AI assistants, this means access to contextual signals such as user location, surrounding geometry, and biometric data. The result is more relevant and timely assistance that adapts to the user’s environment rather than forcing them into a predefined workflow.
Key trends driving this convergence (2026)
- Nous Research’s Bot Mode for Hermes Agent
- Zhipu’s GLM-5.3
- Samsung Health AI
How to build a spatial AI assistant: a step-by-step guide
1. Define the spatial context model
Start by modeling the environment. A simple JSON representation of the scene:
{
"room": "office",
"coordinates": {"x": 3.2, "y": 2.1, "z": 0},
"landmarks": ["whiteboard", "screen", "plant"],
"user_state": {"health": {"bpm": 72, "stress": 0.3}}
}This schema is compatible with most spatial mapping APIs and can be updated in real time with sensor data.
2. Choose a multimodal pipeline
Use a foundation model like GLM-5.3 to process text, depth-maps, and voice inputs. Combine it with a lightweight reasoning engine (e.g., Hermes Agent) to generate actions.
3. Implement Bot Mode for multiple personalities
With Bot Mode you can assign each assistant a distinct specialization:
- Guide Bot
- Coach Bot
- Helper Bot
Example code for initializing a bot via the Hermes API:
from hermes_agent import HermesAgent
agent = HermesAgent(name="GuideBot", personality="informative", license="MIT")
agent.add_capability("spatial_navigate", {"range_meters": 5})
agent.activate_bot_mode()4. Create a space-engineered prompt system
Prompts should include explicit spatial cues to guide the model. An effective template is:
"[Context]: The user is in {room} at coordinates {coords}. {landmarks} are visible. Health status: {health}. Query: What is your primary question?"This engineered prompt reduces ambiguity and improves answer relevance in 3D environments.
Practical example: an office collaboration assistant
Imagine a designer wearing an AR headset while examining a product prototype. The spatial AI assistant:
All three bots are orchestrated through Hermes Agent, share the same spatial context, and respond to the user as a single, coherent voice.
Current trends: from GLM-5.3 to Samsung’s health models
GLM-5.3 enhances the ability to understand spatial coordinates and visual cues, making it easier for assistants to interpret scene geometry. Meanwhile, Samsung’s AI health models transform wearable data into contextual suggestions, a crucial component for assistants operating in time-sensitive environments.
Practical tips and prompt-engineering advice
- Always embed explicit spatial data.Even a small JSON helps the model stay anchored to the scene.
- Use multiple bots to separate concerns.Keeping bots specialized reduces computational overhead.
- Continuously evaluate.Record assistant responses against user expectations and refine the prompt model accordingly.
- Leverage wearable health models.
Conclusion: create truly immersive experiences
Bottom line:use these steps as an operational foundation, adapting tools, policies, and controls to your organization’s real-world context.