Why Big Tech can't quit smart glasses

What It Covers

The Vergecast's David Pierce and Victoria Song evaluate 11 potential smart glasses features — from navigation and camera to AI translation and facial recognition — scoring each one to five to determine why major tech companies remain committed to smart glasses despite persistent technical, privacy, and consumer adoption challenges.

Key Questions Answered

• •Audio/Headphone Mode: Rated five out of five, open-ear speaker mode is the single most-used smart glasses feature because it preserves situational awareness while listening to podcasts or music. Unlike noise-canceling earbuds, glasses allow users to hear ambient sounds — conversations, traffic, household members — making them a practical daily-wear alternative to traditional earbuds.
• •Camera Use Cases: Rated four out of five, the hands-free camera finds genuine utility in three specific contexts: candid home video of children or pets, POV content creation for platforms like TikTok, and accessibility tools like Be My Eyes for low-vision users. Privacy concerns will push some buyers toward camera-free models, splitting the market.
• •Notifications Need AI Filtering: Rated four out of five, notification delivery becomes a killer app only when AI can intelligently triage urgency. Current glasses push every alert indiscriminately. The actual product need is context-aware filtering — suppressing notifications during meetings and surfacing only time-sensitive ones — a problem no manufacturer has solved as of mid-2026.
• •Fitness/Sport-Specific Glasses: Rated three out of five, sport-specific smart glasses like the Meta Oakley Vanguards succeed because buyers accept purpose-built gear without expecting all-day wear. Runners and cyclists benefit from heads-up pace, heart rate, and route data. This single-use-case model sets a lower adoption bar than asking consumers to wear prescription smart glasses all day.
• •AI Features Underperform Outside Demos: Rated two out of five, ambient AI interactions consistently fail in real-world conditions — object identification misses obvious logos, recipe suggestions offer trivial answers, and live translation breaks down in noisy multilingual environments. AI features perform well only in tightly controlled demos, not in the unpredictable conditions of everyday life.