For the 2.2 billion people worldwide affected by visual impairment, navigating a sidewalk, crossing a street, or entering an unfamiliar building remain daily challenges that erode independence and quality of life. Existing assistive systems either lack real-time spatial awareness, depend on cloud processing with prohibitive latency, or rely on GPS data unusable indoors. Most also require devices that draw unwanted attention, a consistently reported barrier to adoption. This work presents a wearable navigation system built on Meta Aria smart glasses, an 85 g platform indistinguishable from regular eyewear, that provides real-time obstacle detection, depth estimation, gaze-aware alert attenuation, and spatial audio feedback within a socially acceptable form factor. A priority-based alert engine with vehicle-first guarantees and gaze-contingent scoring translates complex visual scenes into directional audio cues, reducing information overload while preserving safety-critical warnings. The system achieves real-time operation on a consumer mobile GPU, demonstrating that comprehensive multimodal assistive navigation no longer requires specialized or stigmatizing equipment.