Sabi Raises $50M Seed From Khosla to Put a Brain-to-Agent Interface in a Baseball Cap
Sabi announced a $50 million seed round led by Khosla Ventures, with Accel, Initialized Capital, DST Global, Collaborative Fund, Ascend and former OpenAI chief product officer Kevin Weil participating. Fifty million is a Series B number wearing a seed label, and the product is an ordinary-looking baseball cap. Sensors between one and five millimeters, built around Sabi's own EEG chip, read neural signals through hair with no gel, no surgery and no scalp contact. The company says the chip is more than 50 times as power-efficient as the Texas Instruments ADS1299 front end that most research EEG rigs use, which is what lets the electronics hide inside headwear.
The pitch is specifically about agents, not cursors. Sabi's three interaction modes are imagined speech to text, thought to prompt, where a trained mental command becomes an instruction to an AI assistant such as draft this message or fetch that file, and intent prediction, where the system anticipates your next keystroke or on-screen action and offers it for confirmation. Underneath is what it calls a Brain Foundation Model trained on more than 100,000 hours of labeled non-invasive recordings of people imagining words and performing mental tasks, which it claims is the largest disclosed dataset of its kind. Neural data is encrypted at the sensor and inference runs on encrypted data, so Sabi's servers never see raw signals, and the user controls when the cap records and which apps may receive commands.
The founders are CEO Rahul Chhabra and CTO Atmadeep Banerjee, co-first author of MindEye, the NeurIPS 2023 paper that decoded viewed images from non-invasive brain recordings. The team pulls from Kernel, Apple, Microsoft, Meta, Adobe and Nike, plus Stanford neuroscientists. Vinod Khosla's line is the one that will get quoted: you cannot perform a billion brain surgeries to replace a billion keyboards. The money goes to the dataset, finishing the chip, manufacturing early units for the waitlist, and a prototype demo at CES 2027.
The honest framing is that this is a bet on the input side of the agent stack at a moment when the output side has run far ahead. Agents can write, research and act, and humans still steer them through keyboards and voice. If a non-invasive cap can deliver even a low-bandwidth intent channel reliably, the first customers are not consumers typing faster, they are people running a fleet of agents who want to approve, redirect and stop them without reaching for a screen. That connects straight to the containment conversation of the past 48 hours: a brake is only useful if the human can pull it fast.
Announcement: https://www.globenewswire.com/news-release/2026/10/09/3378310/0/en/sabi-raises-50-million-seed-round-led-by-khosla-ventures-to-put-a-brain-ai-interface-inside-a-baseball-cap.html
Company: https://www.sabi.com/
← Back to all articles
The pitch is specifically about agents, not cursors. Sabi's three interaction modes are imagined speech to text, thought to prompt, where a trained mental command becomes an instruction to an AI assistant such as draft this message or fetch that file, and intent prediction, where the system anticipates your next keystroke or on-screen action and offers it for confirmation. Underneath is what it calls a Brain Foundation Model trained on more than 100,000 hours of labeled non-invasive recordings of people imagining words and performing mental tasks, which it claims is the largest disclosed dataset of its kind. Neural data is encrypted at the sensor and inference runs on encrypted data, so Sabi's servers never see raw signals, and the user controls when the cap records and which apps may receive commands.
The founders are CEO Rahul Chhabra and CTO Atmadeep Banerjee, co-first author of MindEye, the NeurIPS 2023 paper that decoded viewed images from non-invasive brain recordings. The team pulls from Kernel, Apple, Microsoft, Meta, Adobe and Nike, plus Stanford neuroscientists. Vinod Khosla's line is the one that will get quoted: you cannot perform a billion brain surgeries to replace a billion keyboards. The money goes to the dataset, finishing the chip, manufacturing early units for the waitlist, and a prototype demo at CES 2027.
The honest framing is that this is a bet on the input side of the agent stack at a moment when the output side has run far ahead. Agents can write, research and act, and humans still steer them through keyboards and voice. If a non-invasive cap can deliver even a low-bandwidth intent channel reliably, the first customers are not consumers typing faster, they are people running a fleet of agents who want to approve, redirect and stop them without reaching for a screen. That connects straight to the containment conversation of the past 48 hours: a brake is only useful if the human can pull it fast.
Announcement: https://www.globenewswire.com/news-release/2026/10/09/3378310/0/en/sabi-raises-50-million-seed-round-led-by-khosla-ventures-to-put-a-brain-ai-interface-inside-a-baseball-cap.html
Company: https://www.sabi.com/
Comments