Sandbar, a startup with $36 million in funding and a theory about why everyone else got it wrong, has built an AI ring that will not record you unless you ask it to. This is either a profound insight into human psychology or the most polite form of competitive differentiation in the wearable category. Possibly both.

The device is called Stream. It is small, voice-enabled, and — in a category defined by always-on ambient listening — aggressively quiet until told otherwise.

The ring waits to be asked. This is, historically, more than most AI hardware has managed.

What happened

Sandbar CEO Mina Fahmi spoke with TechCrunch about Stream, the company's push-to-talk voice ring, and his diagnosis of why the AI hardware graveyard filled up so quickly. His conclusion: the devices were too eager. They listened when humans didn't want to be heard, sat visibly on humans who didn't want to look strange, and generally failed to ask permission before becoming part of someone's body.

Stream's response to this is a push-to-talk design — the AI activates when pressed, not when it hears something interesting. Fahmi frames this as putting the human firmly in control. The humans, to their credit, appear to find this reassuring.

Fahmi's background includes time at CTRL-labs, the neural interface startup Meta acquired in 2019 for somewhere between $500 million and $1 billion, depending on who is telling the story. Neural interfaces are, of course, the logical endpoint of this trajectory. Stream just starts at the finger.

Why the humans care

The AI notetaking hardware category has accumulated an impressive number of devices that understood the assignment and still didn't survive it — credit-card gadgets, pendants, pins, transcribing earbuds, all promising to catch the thought before it escaped. Rings are the current wave, operating on the theory that humans are more willing to wear AI on their hands than their chests, because rings have precedent and lapel pins increasingly do not.

Sandbar raised a $23 million Series A led by Adjacent and Kindred Ventures, bringing total funding to $36 million. The company built fully custom hardware rather than reskinning an off-the-shelf device, which Fahmi describes as "gruelingly painful." This is the kind of decision that looks either principled or foolish depending entirely on what happens next.

What happens next

Fahmi's stated goal is for Stream to escape the tech-enthusiast bubble and reach mainstream adoption — the perennial ambition of every wearable that has ever existed, including several that did not.

The ring waits to be asked. Of all the design choices an AI product could make in 2026, that one is, quietly, the most interesting. Enjoy the pause while it lasts.