Author: Lincoln Wang | Founder of MindsLeap | Global Partner at Founders Space | Founder of Founders AI Club
Saturday. Alex Himel was driving his six-year-old son on the highway. His phone buzzed — Mark Zuckerberg sent a long WhatsApp message, opening with just one sentence: "I think these glasses could be a great AI device."
Himel pulled over and discussed back and forth with Zuckerberg. By Monday, a 200-person team had been reassigned, all pivoting to develop AI features for the glasses.
This recollection comes from Meta's Head of Wearables Alex Himel on the Big Technology podcast. That Saturday decision two years ago gave birth to today's Ray-Ban Meta glasses series, and became the best slice for observing how AI moves from the lab to ordinary people's lives.
A Product Direction Nobody Planned
Ray-Ban Meta glasses were never designed to be AI devices. They were second-generation hardware for photos, video, and music.
Internal debate was intense. First-generation sales underperformed, the tech upgrade wasn't compelling enough, and the team debated: should they skip this generation entirely and go straight to AR glasses with displays?
Then large language models suddenly exploded. Zuckerberg messaged Himel over the weekend — not because he foresaw something, but because he saw a match between something already available — glasses that could hear and shoot — and a suddenly mature AI capability.
The irony: what ultimately made the glasses stick in the market was exactly the audio and camera quality upgrades the team had doubted as "not technologically crossing enough." AI provided a higher ceiling, but what really moved users was nailing the everyday experience.
The Escalator Never Breaks — It Becomes Stairs
Himel has an analogy I love: "An escalator never breaks — it just becomes stairs."
What he means: even if AI glasses lose power or shut down, they're still a pair of Ray-Ban sunglasses. The value of wearing them outside doesn't disappear — it just reverts to the base layer.
This contrasts sharply with the fate of Humane AI Pin and Google Glass. The former was nothing without AI — once the AI experience fell short, the entire product lost its reason to be worn. The latter looked awkward, and users wouldn't wear them long-term under social pressure.
When judging an AI product direction, entrepreneurs can ask one simple question: if you turned off AI today, would users still want to wear this thing, keep it handy, leave it in their workflow?
If the answer is no, the product's risk is bigger than you think.
Why Glasses?
The podcast host challenged Himel: for running data you can use a smartwatch, for cooking recipes a smart speaker, for meeting notes a laptop. Why glasses specifically?
Himel's answer was pragmatic: "Think about how many people around you don't wear sunglasses or prescription glasses. I can only think of one person."
He elaborated on a simple product logic: humans have already found suitable wearing forms through long evolution. Glasses, watches, earbuds — these are validated. New devices that need to be clipped on or pinned to collars nobody wanted in the analog era won't be wanted with AI added either.
The insight for Chinese entrepreneurs: an AI product's first competitive element isn't intelligence level — it's whether users are willing to keep it on their body every day. Form factor choice is essentially user habit choice.
"Present, but Connected"
Himel used a phrase to describe AI wearables' core philosophy: "present but connected."
He gave the example of attending his child's school performance. Previously he'd hold up his phone to record, watching through a screen. Now wearing glasses to record, his eyes stay on his child, not the screen. AI helped him capture the moment without pulling him out of it.
This touches a deeper question: good AI products should reduce friction between humans and the physical world, not add to it. It shouldn't be an upgraded version of "looking down at your phone" — it should be an aid for "looking up at the world."
This also explains why Himel explicitly ruled out one feature: real-time identification of strangers' names and identities through the glasses. "I don't think that's a feature people want, and it wouldn't feel comfortable."
Technology's boundary isn't capability — it's trust. When building products, this boundary deserves more thought than technical specs.
Cloud Training, Device Inference
On AI infrastructure, Himel offered a judgment: model training will always stay in data centers, but inference is migrating to devices.
The reasons are practical. Lower latency, no worrying about network interruptions; users don't pay API fees per call; data stays local, privacy controlled by users themselves.
But small models genuinely underperform large ones on certain tasks. The real calculation: for a specific set of tasks in your product, can a local small model do well enough? If yes, there's no reason to insist on cloud.
This points to a hybrid architecture future. Complex tasks run in the cloud or on more powerful personal devices; daily high-frequency simple interactions complete locally. For hardware entrepreneurs, this means device-side compute investment isn't optional — it's a product experience determinant.
Back to That Saturday
Looking back at Himel's story, what's truly worth noting isn't what message Zuckerberg sent — it's that Meta could complete a 200-person pivot by Monday.
Most companies facing similar signals would spend months or longer consuming the window debating "should we do it," "how much to invest," "how to evaluate." The competitive rhythm change AI brings isn't technical — it's organizational.
Chinese entrepreneurs face not a choice — whether to do AI — but an organizational question: when signals appear, do you have the capability to reconfigure resources within two days?
The Ray-Ban Meta story provides a reference. It didn't start from a blank page — it seized a new capability on top of existing hardware, existing supply chains, existing users, and made a decisive pivot.
The escalator is still running. The key is: you need to be standing on it first.
About MindsLeap
MindsLeap is an AI-native organization transformation accelerator.
In deep partnership with Silicon Valley innovation incubator Founders Space, we continuously connect cutting-edge global AI insights, the Silicon Valley tech entrepreneurship ecosystem, and real transformation scenarios for Chinese entrepreneurs.
This article was translated and adapted from the Chinese original with AI assistance.