The Subvocal Company, a Y Combinator F26 startup, says it is building smart glasses that let users talk to a computer without speaking aloud: wearers mouth words silently and the glasses decode them to text “at the speed of speech,” according to a “Show HN” post on November 2, 2026 by its founders, Aarush and Arnav. The post drew 837 points and 393 comments on Hacker News. The claims below are the founders’ own, and Lens did not see accuracy figures in the post.

What the founders say

  • How it works: the glasses emit radio waves, “similar to Bluetooth,” that pass through the face. As they cross the tongue, lips and soft palate they are modulated by silent mouthing, and the team trains foundation models to decode those perturbations to text. A later edit says users do not have to move their lips or open their mouth, because the radio signal can see inside the mouth and the system can be fine-tuned on “internal speech,” which the founders describe as like mouthing with the lips closed and “a bit weird to do.”
  • Origin: the founders say they were Georgia Tech computer-science students who tried to recreate MIT’s AlterEgo silent-speech system in 36 hours at Stanford’s TreeHacks in February, using electrodes on the face. They say muscle-signal (EMG) sensing was abandoned because signals change with sweat, interference and placement, and because “no one wants to stick electrodes onto their face.” They then built a radio-frequency sensing stack and glasses with the antennas embedded in the frames.
  • Training data: to handle differences in facial anatomy across users, they say they paid people to read prompts wearing the glasses in booths at their office and scaled up the pretraining data. In a video, a YC partner identified as Dave used the glasses after “only a short 20-sentence calibration.”
  • Availability: they say they will keep working over the next few months to turn the prototype into the product shown on their website and ship in Q1 2027, and have opened preorders for a first batch of 500 units with a $200 discount off the final price. The post does not give the final price.

Reaction on Hacker News

Commenters were mixed. One said talking silently in public would feel awkward, others pointed to open-source hobbyist projects with similar goals, and some asked why glasses are needed rather than a throat microphone, or whether the effort is worth it compared with fast typing. These are individual views, not findings.

Sources and notes

This post is based on the founders’ Show HN post (November 2, 2026), read through Hacker News’s public API, and its top comments. Lens could not read the company’s website, did not watch the demo video, tested no hardware and saw no independent evaluation of accuracy or speed. Lens did not contact the company.

Image: black-framed eyeglasses resting on a wooden desk, by Shixart1985 via Wikimedia Commons, licensed under CC BY 2.0; cropped. Stand-in photo; these are ordinary glasses, not Subvocal’s product, and the photo does not show the events described.