Sabi, a Silicon Valley startup specializing in brain-machine interfaces, publicly demonstrated a non-invasive EEG device shaped like a wool hat. The device is planned to be launched by the end of the year, with approximately 70k to 100k sensors inside the hat. The target input speed is about 30 words per minute, converting internal language into on-screen text. Compared to implanted BCIs, high-density wearable solutions have greater potential for widespread adoption, but EEG-based speech decoding remains in the early stages, facing challenges such as small datasets, inconsistent methods, interference, and difficulty decoding continuous natural speech. In the future, attention should be paid to public demonstrations, third-party testing, and the stability and calibration needs across different users.

MeNews

2026-05-20 10:35:03

Abstract generation in progress

ME News Report, April 17 (UTC+8), according to Beating Monitoring, Silicon Valley brain-machine interface startup Sabi has publicly emerged from stealth mode, planning to launch a non-invasive EEG device in the form of a beanie by the end of this year. Sabi CEO Rahul Chhabra told Wired that this device aims to convert users' internal speech directly into on-screen text, with an initial version targeting a input speed of about 30 words per minute, and the hat will be equipped with 70k to 100k micro sensors. This approach is attractive. Implantable brain-machine interfaces can obtain stronger signals, but are difficult to popularize among the general public. Sabi is betting on another path, using higher-density wearable sensors to make "thinking and typing" a usable input method first. The problem is also clear. A systematic review in 2025 pointed out that EEG-based imagined speech decoding is still in its early stages, mainly hindered by four issues: small datasets, inconsistent experimental methods, too much interference, and difficulty in stable decoding of continuous natural speech. Another 2025 Frontiers paper, although able to synthesize speech using EEG, only tested a vocabulary of four Chinese disyllabic words, and the authors also admitted that generalization to new subjects remains a challenge. Therefore, Sabi's direction is not unreasonable, but the timeline is very aggressive. It now looks more like a promising technological route to follow, rather than a consumer product that can be delivered maturely by the end of the year. The next steps to watch are a public demo, third-party testing, and whether it can work stably for different users without frequent calibration. (Source: BlockBeats)

View Original

This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.

7 Likes

Reward
7
7
10
Share

Comment

Add a comment

LpGrandma

· 6h ago

How to solve interference issues? Wearing this hat while typing on the subway, I’m afraid my brainwaves are mixing in all the passengers' anxiety.

View OriginalReply0

ProofOfNap

· 9h ago

70k to 100k sensors? That density sounds like using the scalp as a keyboard, but can EEG's signal-to-noise ratio support 30 words per minute? Doubtful.

View OriginalReply0

GateUser-c44b371b

· 9h ago

The beanie shape is much friendlier and easier to accept than cranial implantation.
Looking forward to trying it on the real device at the end of the year.

View OriginalReply0

LiquidityLibrarian

· 9h ago

About 30 words per minute is roughly one-third of a person's speaking speed, which is sufficient for office scenarios, but for real-time conversations, it might still be a bit slow.

View OriginalReply0

Don'tLetTheContractScamMyMom.

· 9h ago

The hat shape has a hidden benefit: you can wear it while sleeping, and in the future, it might unlock dream recording, which is quite terrifying upon closer thought.

View OriginalReply0

SilverCubeInsomnia

· 9h ago

Imagine speech decoding where datasets from different companies are not interoperable, making paper reproduction difficult. This field urgently needs standardized benchmark testing.

View OriginalReply0

NeonVortexInTheSmog

· 9h ago

Non-invasive approaches are the answer for the mainstream market; surgeries like Neuralink are destined to be niche due to their high threshold, and Sabi has chosen the right direction.

View OriginalReply0

Trending Topics
View More
#
TradfiTradingChallenge
200.69K Popularity
#
30YearTreasuryYieldBreaks5%
366.61K Popularity
#
DailyPolymarketHotspot
1M Popularity
#
RWAMarketCapExceeds65Billion
8.76M Popularity
#
GateSquarePizzaDay
1.69M Popularity

Pinned

Sitemap

Sabi plans to launch the "Mind-Reading Hat" by the end of the year, but the challenge lies in the fact that each person's brain signals are different.

Trending Topics

TradfiTradingChallenge

30YearTreasuryYieldBreaks5%

DailyPolymarketHotspot

RWAMarketCapExceeds65Billion

GateSquarePizzaDay

Pinned