My working name for Brim was "half-empty tennis cap". It's a visor (if you want the word the shops use), and I wanted it to carry an AI agent.
That sounds like a joke. I think it's the most practical idea in the whole concept: put the agent only where your head has already said yes.
Here's the reasoning, the research that pushed back, and the one part that has to wait.
Every part has to live rent-free
I call the rule "rent-free parts". Nothing goes on your head unless that spot already has a job.
A visor passes the test well. The band already wraps the head to hold it on. The brim already sits out front, shading your eyes. Neither touches your face.
The electronics would only move into space that's already paid for. The strap would get motors, the brim a camera at its front edge, and a short arm off the strap would carry an earpiece to one ear.
Prescription glasses? Keep them. Nothing on Brim would sit on your nose.
Display glasses make the opposite trade. They put the screen exactly where your eyes are, then have to earn that spot every minute you wear them.
Meta's Ray-Ban Display fits a 600 by 600 display into the right lens, which is impressive (and a lot to ask of a face).
Rent-free isn't free, though. A visor has no crown, and Brim's brim would be short on purpose.
A 1992 study of sun hats by Diffey and Cheeseman found a brim has to be wider than 7.5 cm to give the nose and cheeks reasonable protection. Mine would be about 6 cm.
So Brim would shade your eyes, not your whole face, and sunscreen keeps its job.
My test for any new part: "Is this spot already on my head for another reason?"
The strap taps one motor at a time
The strap was inspired by Walnut's EEG headband. (Walnut is a neural-interface startup founded in Stanford, California.) Their band listens to your head. I wanted mine to talk back.
The tempting design draws shapes with buzzes, like an arrow sliding left across your forehead. It looks great in a sketch.
Skin disagrees. In a 2012 study by Dobrzynski and colleagues, 22 people wore 12 coin motors around the head. With one motor buzzing, they counted it right 94% of the time. With two, 40%. With five, 5%.
Brim would play one motor at a time instead. "Turn left" would be a few taps walking toward your left temple.
"Muse has something" would be a double tap on the forehead. The same study found a buzz hardest to place at the front and back of the head, so the forehead gets the one cue where precision doesn't matter.
Hair gets a vote, too. A 2015 study by Myles and colleagues found dense hair may slightly dampen vibration at the head's least sensitive spots, so each motor's strength would need to be adjustable.
That's a good kind of constraint. It would make the design simpler and the cues easier to tell apart.
The research also points somewhere bigger than a running gadget. Dobrzynski's study opens with visually impaired people, and taps leave the ears free for traffic.
A strap that points the way without talking over the street could matter more to a blind pedestrian than to anyone else wearing Brim. It still wouldn't replace a cane.
Which Muse, though?
Brim would run on Muse, Meta's personal AI agent, through the open-source Muse Gadget SDK. And its strap borrows from an EEG headband.
Search "Muse headband", though, and you'll find a different Muse: InteraXon's EEG headband, which has nothing to do with Meta's agent.
So an AI wearable shaped like a headband, inspired by an EEG headband, sits right next to an EEG headband called Muse.
The case study says which Muse I mean. (The agent. The EEG headband would be a very different project.)
The SDK already covers a lot. Pairing, push-to-talk and custom commands are there.
One detail matters more than it looks: Muse answers in text. Spoken replies need a text-to-speech step the SDK leaves to you, so the earpiece needs a speech path I'd have to build.
The wire is a separate call. The chip I'd use can't stream audio to ordinary earbuds.
The hologram has to wait its turn
The wildest idea is the least buildable one: a screen projected down from the front of the brim, floating in front of your eyes.
The bad news comes first. A hologram can't float in open air, because you only see light that scatters off something or is steered into your eye.
So what would the "hologram" be? A holographic film on a small clear panel that flips down under the brim, steering a projector's light into your eye.
North's Focals glasses used that trick inside a lens, at about 300 by 300 pixels.
Then the harder news. Glasses sit on your nose and ears, locked to your eyes. A visor rides on your forehead, so raise your eyebrows and the brim moves.
A display that moves when you frown is a display you stop trusting. That's why it's the last thing I'd build, not the first.
Where this breaks
None of this is built. Every number is a target or comes from someone else's study.
The first prototype I'd make is just the strap and a small ESP32 board, worn for a week before any camera goes near it.
A visor isn't for everyone, either. It's a sun-and-sidewalk object, not an office one, and hair, helmets and other hats all compete for the same spot.
I'd rather fit one place well than everywhere badly.
Everything on Brim would live rent-free, except the hologram. It can move in once it can pay. The full concept, with the strap demo, is in the Brim case study.