The takeaway: Typing may be the dominant way humans interact with computers, but it is bound by keyboards, touchscreens, and voice commands. A growing group of engineers is now working to replace it with something more direct: human thought.
Silicon Valley startup Sabi is the latest entrant to suggest using the brain as an interface device. The company is developing a noninvasive device that translates internal speech into text. Rather than relying on implanted hardware, Sabi is building a wearable device – initially in the form of a beanie, with a baseball cap version coming later – that captures neural activity and converts it into words displayed on a screen.
The approach places Sabi in a different category from companies like Neuralink, which are pursuing surgically implanted systems aimed primarily at patients with severe motor impairments. Sabi is targeting broader, everyday use, betting that wearability – not invasiveness – will determine whether the technology reaches scale.
"The biggest and baddest application of BCI is if you can talk to your computer by thinking about it," Vinod Khosla, founder of Khosla Ventures and an early investor in the company, told Wired. "If you're going to have a billion people use BCI for access to their computers every day, it can't be invasive."
Sabi's system uses electroencephalography (EEG), a technique that records electrical activity in the brain through sensors placed on the scalp. The medical and research industries have used EEG for decades, and recent advances have enabled the decoding of limited forms of imagined speech. The challenge is that these signals are weak and diffuse when measured noninvasively, especially compared to the high-fidelity data captured by implanted electrodes.
Sabi's answer is scale. While conventional EEG systems may use anywhere from a dozen to a few hundred sensors, the company is designing a device with tens of thousands. It claims the cap will include between 70,000 and 100,000 tiny sensors, dramatically increasing spatial resolution.
"Given that high-density sensing, it pinpoints exactly what and where neural activity is happening. We use that information to get much more reliable data to decode what a person is thinking," says CEO Rahul Chhabra.
Even with denser sensing, decoding internal speech remains a complex problem. Neural patterns associated with language vary not only across individuals but also within the same person over time. Factors such as fatigue, attention, and context can subtly shift how the brain encodes intended words, complicating real-time interpretation.
To address this variability, Sabi is developing what it describes as a brain foundation model – an AI system trained on large-scale neural datasets to identify common patterns linked to inner speech. The company says it has already collected roughly 100,000 hours of brain data from 100 volunteers. Unlike models tailored to a single user, this system can generalize across many individuals, a requirement for any consumer-facing product.
Performance remains an open question. Sabi is targeting an initial typing speed of about 30 words per minute, which is slower than most people type but expected to improve over time. Chhabra expects accuracy and speed to improve as users spend more time with the device.
However, technical performance is only part of the challenge; everyday usability may be harder. Many existing BCI systems require calibration before each session, a process that limits their practicality outside controlled environments. For consumer adoption, that friction has to disappear.
"These devices are going to have to be ready to go out of the box," says JoJo Platt, an independent neurotechnology consultant. "They're going to have to conform to me rather than me conforming to it."
Form factor is another unresolved issue. Wearable devices must balance sensor density with comfort and discretion, especially when intended for daily use. Even in medical contexts, users tend to prefer unobtrusive or invisible systems, a factor that has influenced the design of both implanted and wearable alternatives across the industry.
At the same time, the shift from physical input to neural data introduces a new category of risk. Brain signals, by definition, carry highly sensitive information, raising concerns about how that data is stored, processed, and secured. Sabi says its system encrypts data end-to-end when transmitted to the cloud and allows AI models to train on encrypted datasets rather than raw signals. The company is also working with external neurosecurity experts to audit its infrastructure.
"We need to recognize that neural data is the most private kind of data that a person could possibly have," Chhabra says. "Not treating it with care would just be unfair."
Whether noninvasive BCIs can achieve the fidelity needed for seamless communication remains uncertain. For now, Sabi's bet is clear: skip surgery, pack as many sensors as possible into ordinary headwear, and use large AI models to turn noisy brain signals into text that anyone can type just by thinking.