On October 9, Palo Alto startup Sabi raised a $50 million seed round, led by Khosla Ventures, to build a wearable that turns a thought into a prompt for an AI assistant. Forbes puts the valuation around $600 million. Accel, Initialized Capital, DST Global, Collaborative Fund, Ascend, and former OpenAI chief product officer Kevin Weil joined the round.
You can't buy one. Sabi hasn't published a price, and nothing is on sale. Sabi's announcement puts the device in a baseball cap, with an early demo at CES 2027. CryptoBriefing puts it in a beanie the company aims to ship by the end of 2026, with the baseball cap as another form in the pipeline. Until one of those versions has a price, you're looking at a funding story with two timelines.
Hold that gap. A beanie with an end-of-2026 ship target and a baseball cap with a CES 2027 demo are different objects on different clocks, and neither has a sticker. The $50 million seed and the valuation around $600 million are the figures you can check today. The decode rate, the day you could buy one, and the price are still ahead of you.
The hat reads the brain's electrical activity through hair, with no gel and no surgery, and the target is about 30 words a minute. Neural data is meant to be end-to-end encrypted, with outside audits of the system. CEO Rahul Chhabra and CTO Atmadeep Banerjee started the company.
The signal has to cross the skull
Implants such as Neuralink sit on the other side of this bet. They ask for surgery so the sensors can sit close to the tissue. Sabi is betting you'll only try a brain interface if it stays a hat you can take off. Skull, scalp, and hair blur the detail an implant can pick up. The company is stacking a dense sensor array and a model that's supposed to fill in what the electrodes miss. A hat that looks ordinary still has to hear a thought clearly enough that an assistant acts on the right request.
Wait for a measured decode rate on a new wearer, shown in public, and notice whether the CES demo is that measurement or a scripted lab run. Treat about 30 words a minute as a target until you can see it work on someone the model hasn't seen. Wait for a price, and for a plain account of how long calibration takes, before this goes anywhere near a buy list. If the next update still can't tell you whether you're buying a beanie in 2026 or watching a cap demo in 2027, leave it off that list.
