Google's Smart Glasses Are Back. The Second Wave Will Tell Us What AI's Form Factor Actually Is.

The line everyone uses about Google Glass is that it was ten years too early. Looking at the May 2026 I/O announcements, that line was wrong twice. Glass was not ten years too early, and what is happening now is not just Glass with better hardware. The second wave of smart glasses is the first form factor that builds around the existence of a competent AI agent rather than bolted on to a 2013 software stack. That makes the design problem fundamentally different, and the next twelve months will tell us which interaction patterns survive contact with real users.

What Google announced and what is shipping

At I/O 2026, Google previewed "Intelligent eyewear with Gemini" launching in two phases. Audio glasses arrive this fall. Display glasses follow later. Launch partners are Samsung, Warby Parker, and Gentle Monster. iOS and Android compatibility is named. Geography is not committed in Google's post but the coverage suggests US-first.

The feature set Google has confirmed: "Just say 'Hey Google' or tap the side of the frame to instantly access Gemini," paired with "over-ear speakers," photo and video capture, and "Ask Gemini about anything you see" for ambient camera context. Google explicitly named the Warby Parker and Gentle Monster partnerships as a fashion-first move. Android president Sameer Samat told Android Authority the lesson from Glass was that "fashion comes first, technology comes second".

That fashion-first framing is the second tell about how different this wave is. Glass tried to make the technology socially acceptable. The 2026 wave is trying to make the technology socially desirable by partnering with brands people already wanted to wear before any AI was involved.

Meta shipped first. The Meta Ray-Ban Display launched September 30, 2025, US-only at $799 with the Neural Band, and Meta paused the international rollout to the UK, France, Italy, and Canada at CES 2026 due to demand and inventory constraints. Orion remains a research prototype. The lighter Artemis variant is targeted for 2027.

Snap opened pre-orders for its sixth-generation Specs, its first consumer AR glasses, at $2,195, shipping this fall in the US, UK, and France. Standalone (no puck), with OpenAI and Gemini integrations on Snap OS 2.0.

Apple is the wildcard. Bloomberg's Mark Gurman has reported a late-2026 preview and 2027 retail launch for display-free smart glasses (codename N50), with Vision Pro 2 deprioritized. No official Apple announcement. Treat as rumor-tier until Apple says something on the record.

So by Christmas 2026 you can expect at least three major shipping smart-glasses products from Meta, Google, and Snap. Apple in 2027 if Gurman is right. That is not a Glass replay. That is a real category launch.

What changed since 2013?

The single biggest difference is that smart glasses in 2013 were a phone accessory. Smart glasses in 2026 are an AI agent surface that happens to be on your face. That is not a hardware difference. It is a software difference, and it is what makes the form factor matter at all.

Glass in 2013 ran Android apps in a tiny display. You used it to check notifications, get walking directions, and take videos. There was no contextual model on the other end of the camera. Looking at a restaurant on the way home from work did not produce a useful interaction, because the system on the other side could not interpret a restaurant or know which kind of restaurant you wanted.

Glasses in 2026 ship into a different world. The camera input is fed to a frontier model. The voice input is parsed by a frontier model. The output back to you is a frontier model's response, which means the interaction can be genuinely useful in ambient conditions that were impossible in 2013. "What is this plant" is a one-shot question now. So is "summarize what just happened in this meeting I walked into late." So is "is this restaurant any good and what should I get."

The Glass-era critique that smart glasses were a solution looking for a problem applied to a software stack that could not interpret context. Once the software can interpret context, the critique stops working. The remaining open questions are about social acceptance, battery life, fashion-credibility, and privacy posture, not about whether the product has anything to do.

What does "fashion-first" actually solve?

The Glass failure had three layers. The hardware was visibly weird. The use cases were thin. The privacy posture was openly hostile to the people around the wearer. The 2026 wave is targeting all three.

On hardware visibility, the partnerships with Warby Parker, Gentle Monster, and Ray-Ban are doing the work. The Meta Ray-Ban Display looks like Ray-Bans. The Warby Parker version of Google's audio glasses will look like Warby Parkers. That changes the social cost of wearing the device from "you are obviously a tech person in a meeting with non-tech people" to "you are wearing glasses." A meaningful shift, and one Glass never made.

On use cases, the answer is the AI agent. Whether a frontier model actually does useful things on demand through a wearable camera-and-mic surface is the open question. The Meta Ray-Ban Display has been in the field for about ten months now, and the early reviews suggest the use cases are real but narrow: hands-free music, hands-free messaging, occasional ambient queries. Not life-changing, but durable enough that people are continuing to wear them.

On privacy, the 2026 cohort has done more than 2013 did, and probably not enough. Strings leaked from Google's Android XR Glasses companion app describe an on-device "conversation detection" pipeline that does not send raw audio, images, or conversation data off the glasses. That is the right architectural answer. Whether it is actually how the product behaves in practice, and whether the people in the room with the wearer trust that it is, is a longer conversation. Glass's privacy failure was social, not technical, and the 2026 wave has to win the social battle, not just the technical one.

Why does the form factor matter for product design now?

If you are designing a product that has any AI surface, the interaction patterns that get baked in over the next eighteen months are the patterns the next decade of consumer AI is built around. Glasses are interesting not because they will replace phones, but because they will change what we consider a normal interaction with an AI.

Three patterns are about to harden, and which ones harden is worth paying attention to even if you have no intention of shipping a wearable.

The "ask the room" pattern. Voice-first, hands-free, camera-aware. The traveler asking their glasses about the restaurant in front of them is a natural use case. So is the operator asking about the equipment they are looking at. So is the parent asking about the playground they are watching. This pattern is hard to do well with a phone because the phone is in your hand. Glasses make the camera-aware ambient question native.

The "narrate the moment" pattern. Ambient capture that the user can interrogate later. "What did I just walk into" works in a way it cannot through a phone screen. This is the pattern most likely to be controversial, because the line between useful memory aid and ambient surveillance is thin.

The "agent at your shoulder" pattern. A persistent agent that does not have to be summoned, that can interrupt the wearer with a small note when useful. This pattern is the one no current product has cracked. Whoever ships an interrupt-balance that does not annoy the user wins a meaningful share of the form-factor's mind for the next decade.

For an app or product team, even one with no wearable in the roadmap, the patterns that become normalized in the next eighteen months are the patterns your users will eventually expect everywhere. The transition to mobile in 2007 to 2010 had this shape. So did the transition to voice in 2014 to 2017. The transition to glasses-mediated agentic AI is the third one, and it is happening this year.

How does this connect to the AI strategy work you should already be doing?

If you have read Google's I/O 2026 OS-Layer Bet, the glasses launch is the natural extension of that thesis. The OS layer goes on your face. The agent surface is the camera-mic pair. The cross-app context becomes cross-physical-environment context. The same orchestration logic that lets Spark act across apps will eventually act across what you are looking at.

The Vaib position on this is that 2026 hardware will not be the year glasses become the default consumer device. Meta has shipped, and the numbers are good but not iPhone-scale. Google and Snap will ship and add to the category, not redefine it. The category becomes interesting in 2027 to 2028 when the hardware costs drop, the battery improves, the social acceptance compounds, and the interaction patterns settle.

The strategic question for any business now is whether your product surface will translate. If your product is fundamentally a screen interaction, you have at least three more years before glasses become a relevant competitive consideration. If your product is fundamentally a moment-of-decision interaction (where the user has a question they want answered right now in the context they are in), glasses become a relevant surface much sooner.

What do I bet happens by the end of 2027?

Three predictions, each with a clear way I could be wrong.

By December 2027, at least one display-class smart-glasses product (Meta Ray-Ban Display, Google's display glasses, or Snap's Specs) will report unit sales above three million globally. The bar matters here: Glass never crossed half a million, and while Meta's camera-and-audio Ray-Ban line passed a million units in 2024 and, per EssilorLuxottica, sold roughly 7 million smart glasses in 2025 alone, the display category is still unproven. Display glasses graduating from niche to small-but-real is the actual test of the second wave. If by December 2027 no display product has crossed three million, the category is stalling and the 2013 critique was right after all, just on a longer timeline.

By the same date, at least one of the three "ask the room," "narrate the moment," or "agent at your shoulder" interaction patterns will have a genuine product-market fit moment, with a known consumer use case and reliable retention. My bet is on the ask-the-room pattern: travel, food, and outdoor use cases are the most natural fit, and AI capability is now high enough to make the answer useful instead of frustrating. If by the end of 2027 the patterns are still feeling speculative, the form factor needs more time than I am projecting.

Third, and most contrarian: at least one company will pull its smart-glasses product within the eighteen-month window due to a privacy-related incident or regulatory action. The privacy posture is the soft underbelly of this entire category, and the wearables industry has not yet had its post-Cambridge-Analytica moment. Whoever has the worst on-device privacy story when the moment comes will pay the price, and the post-incident posture of the category will be considerably more conservative than the pre-incident posture.

The smart-glasses question is not "will it work this time." It is "which patterns survive contact with real users." That answer determines a meaningful chunk of how the next decade of consumer AI gets designed. Worth watching, even if you have no intention of putting anything on your face.