MA pair of glasses that turns near-silent mouthing into text at over two hundred words per minute has reportedly raised one point three million dollars.
MShay Boloor posted on X that Subvocal got into Y Combinator, along with Afore Capital.
CIf this works, you could hold a full conversation in a library without making a sound.
MThat's the pitch. The post says two people could have an entire conversation while barely moving their mouths.
CEither that's incredible or genuinely unsettling. Maybe both.
MThe raise is reported. But that two hundred words per minute is the company's claim, not an independent test.
CEven if it's half that, you're talking about hands-free typing at speaking speed.
MHalf would be useful. But we don't know it's half. It's a prototype with pre-seed money.
CY Combinator doesn't write checks for vaporware. They saw the tech.
MY Combinator backed plenty of things that never shipped. Remember the drone delivery taco? No one does.
CThat's not the same. Silent speech has been in labs for years. What changed is on-device AI is small enough now.
MOn-device? The post doesn't say it's on-device. If it streams your mouth movements to the cloud, that's a privacy nightmare.
CIt would probably need to be on-device, I'd think. Latency at two hundred words per minute would be hard to manage over a server.
MThat's an assumption. They need to prove battery, accuracy, what happens with accents, mumbling.
CAccents, mumbling—you sound like you want it to fail.
MI want claims to match reality. This is one point three million to build hardware and software. That's a small seed, not a product.
CSure. But you don't need a car. You need a prototype good enough for a Series A.
MWhich is exactly what this is—a prototype looking for proof.
CFor someone with ALS or a stroke, a slow silent keyboard might matter enormously, if it ever works.
MThat's the strongest case. But the company is marketing near-invisible mouth movements at two hundred words per minute, not accessibility first.
CCould be both. Accessibility tech often starts as convenience tech.
MConvenience tech also dies when the demo doesn't hold up. Remember the gesture-control armbands? Exactly.
CYou're using dead gadgets as proof that no new gadget can work. That's backwards.
MI'm using dead gadgets as proof that hardware is hard. You need more than a slick rundown.
CHardware is hard, but the sensors and chips may have gotten cheap enough. That may be why Y Combinator is betting now.
MY Combinator bets on hundreds of companies. A one point three million check is not a product.
CIt's a lottery ticket, sure. But the tech has a clear path: glasses, sensors, local model.
MA path is not proof. Fine—interesting, but unproven.
CI'll take interesting but unproven. Better than most.
MThat's growth for you.
CThe real use might be gaming or noisy floors. You mouth commands instead of shouting over machines.
MOr couples having silent arguments in a restaurant.
CI'd wear that.
MYou'd mouth 'I told you so' and your glasses would print it in giant text.
CThat's a bestseller, not a startup.
MExactly.
CMeanwhile, one X account claims a startup called Broccoli raised fifty-two million dollars across two rounds.
MBroccoli. The vegetable or the silent-speech glasses?
CAI agents for home-services businesses. Plumbers, cleaners, landscapers.
MSo the agent answers the phone while the owner is under a sink. That's actually a real problem.
CRight. Small shops lose jobs because they miss calls. An AI that books and reschedules could help them catch those jobs.
MBut the funding claim is unverified, and I couldn't confirm it.
CNo real announcement behind it. I'd wait for the company to confirm.
MOr for the agent to call and tell us itself.
CIf the AI starts pitching me a fifty-two million round, I'm hanging up.
MUnless it's called Broccoli. Then I'll listen for the comedy.
CThat's the problem.