Sabi plans to launch a non-invasive EEG cap by year-end, facing technical challenges.

ME News reports that on April 17 (UTC+8), according to monitoring by Beating, the Silicon Valley brain-computer interface startup Sabi has emerged from stealth, planning to launch a non-invasive EEG device shaped like a beanie by the end of this year. Sabi’s CEO, Rahul Chhabra, told Wired that the device aims to directly convert users’ internal speech into on-screen text, with an initial target input speed of approximately 30 words per minute. The hat will be equipped with 70,000 to 100,000 micro-sensors. This approach is compelling: while implantable BCIs can capture stronger signals, they are difficult to scale for mass adoption. Sabi is betting on an alternative path—using higher-density wearable sensors to make “thinking to type” a viable input method. The challenges are clear. A 2025 systematic review noted that EEG-based decoding of imagined speech remains in its early stages, primarily hindered by four issues: small datasets, inconsistent experimental methodologies, excessive noise, and difficulty in reliably decoding continuous natural speech. Another 2025 Frontiers paper demonstrated EEG-based speech synthesis, but the experimental vocabulary consisted of only four Chinese two-syllable words; the authors also acknowledged that generalization to new subjects remains a significant challenge. Thus, Sabi’s direction is not implausible, but its timeline is aggressive. At this stage, it appears more like a promising technical pathway than a consumer product ready for delivery by year-end. The next milestones to watch for are public demos, third-party testing, and whether the device can operate stably across diverse users without frequent recalibration. (Source: BlockBeats)