Sabi

Sabi is a California-based brain-computer interface (BCI) startup developing a wearable cap, marketed as the Sabi Cap, that is designed to read a wearer's thoughts and attempted speech and convert them into text. The device is a beanie (with a baseball-cap version also under development) that works wirelessly and is intended to let users operate computers and direct AI agents without typing or clicking.

The company's stated approach has three parts: brain imaging using custom neuroimaging sensors with a 1mm footprint powered by proprietary ASICs; large-scale collection of neural decoding data; and a "Brain Foundation Model" that maps brain signals to thoughts. Sabi describes its method as based on electroencephalography, recording electrical activity through the scalp, and says it compensates for signal attenuation through bone and skin by using far more sensors than conventional EEG systems — on the order of 70,000 to 100,000 per device, compared with a dozen to a few hundred in typical EEG hardware. Its decoding model has been trained on roughly 100,000 hours of neural data gathered from about 100 volunteers, and the company states that data from the device is kept private.

Sabi positions its non-invasive, wearable approach against surgically implanted BCIs such as Neuralink. Initial typing performance is described as about 30 words per minute, which the company expects to improve with continued use. The product was expected to launch by the end of 2026.

Founding story

Sabi was co-founded by Rahul Chhabra, a graduate of BITS Pilani in Rajasthan, India.

Business model

Sabi develops a consumer hardware product — a beanie or cap containing its own neuroimaging sensors and ASICs — paired with in-house decoding models. The company plans to sell the device to end users, with a waitlist collected on its website ahead of launch; pricing has not been disclosed.

Traction

Sabi reports having collected about 100,000 hours of brain data from 100 volunteers to train its decoding model, and reports an expected initial typing speed of roughly 30 words per minute. The product had not shipped as of April 2026, with a website waitlist open and launch anticipated by the end of that year.

Latest developments

In April 2026, WIRED profiled Sabi's thought-to-text beanie, prompting wider coverage. At that time the company disclosed that its decoding model had been trained on about 100,000 hours of neural data from 100 volunteers, that initial typing speed would be around 30 words per minute, that a baseball-cap form factor was also in development, and that the product would likely launch by the end of the year, with pricing not yet announced.