The AI hardware graveyard just got a counterexample worth watching.
The Summary
- Pocket, an AI notetaker that clips onto smartphones, hit a $100M run rate eight months post-launch with 200,000+ devices sold and just 12 employees
- Run rate isn't actual revenue. It's current pace projected forward, but 50% month-over-month growth with subscription income is real traction
- The key insight: they started with an app, pivoted to hardware when users demanded it, and chose the form factor that actually fits workflow instead of sci-fi aesthetics
The Signal
Pocket did something most AI hardware startups refuse to do. They listened when the market told them software wasn't enough. Doctors and salespeople kept asking for a device that didn't require unlocking a phone between back-to-back appointments. So cofounders Akshay Narisetti and Gabriel Dymowski, who met at a Sam Altman-backed hacker house, built a thin recorder that snaps onto the back of a smartphone.
The form factor matters more than the tech press wants to admit. Humane's $700 AI pin died in 2024 because it tried to replace your phone. Pocket survives because it enhances the device you already carry. One button. No friction. No future-of-computing manifesto required.
"The graveyard of AI hardware is full of devices that prioritized vision over workflow."
The unit economics tell a clearer story than the run rate hype. Two hundred thousand devices sold in eight months with a team of twelve means this thing has actual demand, not just VC-subsidized curiosity purchases. The subscription revenue Narisetti mentions suggests recurring value, not one-time novelty buying. If users keep paying monthly, the device is doing a job they can't get done another way.
Key traction signals:
- 50% monthly growth rate sustained over eight months
- Subscription model generating recurring revenue at scale
- 16,667 devices sold per employee (200k÷12), absurdly high efficiency
But run rate is a projection, not a fact. A $100M run rate means if current monthly revenue stays constant for 12 months, they'd hit $100M. The claim of reaching $1B run rate in two years assumes that 50% monthly growth continues, which it won't. Growth rates compress as you scale. The real test is whether Pocket has found product-market fit strong enough to survive the inevitable plateau.
The Y Combinator pedigree and Altman connection matter less than the actual use case. Busy professionals who talk for a living, doctors who need visit notes, salespeople who need call summaries, need this to work seamlessly or they'll stop using it. The fact that Pocket tested glasses and pendants before settling on a phone attachment shows they're optimizing for adoption, not press coverage.
The Implication
If Pocket's numbers hold, it proves there's a narrow path for AI hardware that augments existing behavior instead of trying to replace it. Watch whether the subscription revenue base grows faster than device sales. That's the signal that this is a tool, not a gadget. And watch whether enterprise buyers start ordering in bulk. If corporate sales teams and hospital systems start issuing these to staff, Pocket might actually be building infrastructure instead of selling novelty.
The broader lesson: the agent economy doesn't need new wearables. It needs better interfaces to the devices we already use. Pocket figured that out. Most AI hardware startups still haven't.