VPS for Ray-Ban Meta AI Glasses

Ray-Ban Meta AI glasses are launching with Meta AI built in. We're preparing VPS integration for the next generation.

Glasses vs Headset Constraints

Orders of magnitude harder.

Glasses VPS Architecture

Can't run Quest VPS architecture on glasses. Complete rethink:

Sparse queries: Once per minute, not continuous Cloud-heavy: Minimal on-device processing Semantic focus: "Where am I?" not "Exactly where am I?" Multi-modal: GPS + WiFi + visual combined

Different product, different architecture.

Privacy Amplified

Glasses worn in public continuously raise privacy stakes:

• Bystanders don't know when cameras active
• Social norms around glasses different than headsets
• Always-on means more potential capture

Even more stringent privacy requirements:

• Extremely limited data retention
• Clear indicators during any capture
• Location only when user explicitly requests

Use Cases

Different from Quest:

• Contextual AI: "What building is this?" → VPS identifies location → AI provides info
• Navigation: Step-by-step walking directions with audio
• Memory: "Remember I parked here" with visual confirmation
• Social: Share location with context ("I'm at this café")

Less about AR overlay, more about AI context.

Technical Approach

Ultra-efficient features: New network architecture targeting 5mW inference Compressed queries: Send minimal data for cloud matching Predictive localization: Guess location before query completes Graceful degradation: GPS fallback is acceptable

Timeline

Ray-Ban Meta launching 2023 without VPS. Next generation (2024+) targeted for VPS integration.

Using this generation to:

• Understand real sensor characteristics
• Collect training data (consented)
• Refine ultra-low-power algorithms
• Build glasses-specific maps

Preparing for the future where glasses have spatial awareness.