the wand · blog · press · about · home

Three problems. One thing in the hand.

11 September 2026 · John Usher · Perspective

Every screen we have added since the phone has arrived without a keyboard, and the one keyboard we do have is terrible. Nobody appears especially troubled by this. I have been troubled by it for some years, and this is the long version of why.

The elderly are left out of the family chat

An elderly woman sits in an armchair with a tablet on a table in front of her. Inset over her lap, a WhatsApp group called Family chat: five replies between 13:58 and 14:20, the header reading 'typing…', and a message box holding a line that was started and never sent.
The photograph is ours, made with AI. The conversation is the point: five people answered inside the hour, and one is still typing.

A photograph lands in the family group. Everyone answers within the hour, except the one who writes by hand.

Not for want of an opinion. The opinion exists. It is being entered one letter at a time, with a finger, and by the time it arrives the conversation has moved on to somebody’s boiler.

One in six of us will be over sixty by 2030. That is the World Health Organization’s number, and it is not a niche. It is the fastest-growing group of people who own televisions, own phones, and have something to say in the family chat.

The thing to be careful about here is the diagnosis. It is tempting, and wrong, to say that older hands cannot do this. We went and measured it, across 354 people, and they can: the over-sixties are read as well as the under-forties. What reached them late was not the hand. It was the keyboard — which arrived in their fifties, on glass, at the size of a postage stamp.

Phone keyboards are too small for the hands that use them

A phone's on-screen QWERTY keyboard drawn at the size it appears, with a red circle the size of a thumb tip laid over it. The circle covers parts of six keys across two rows.
A thumb tip, to scale, against the keys it is expected to hit.

Thumbs on glass, in 2026, on a layout drawn up in 1874 for the Sholes and Glidden Type-Writer — a machine with a bell on it.

Every other part of the phone has been reinvented twice over since. The camera, the radio, the screen, the silicon, the battery chemistry. The keyboard is a picture of a Victorian machine, shrunk until the keys are smaller than the thumb that hits them.

Which is why autocorrect exists. Autocorrect is not a feature. It is a patch over a hit-rate problem, and everybody knows it, and we have collectively agreed not to mention it.

There is no good text interface for smart glasses

A pair of smart glasses on a display stand.
Three million pairs with a display this year. On almost none of them can you write a word.

You read the message on the glasses. Then you get your phone out to reply, which does rather give the game away.

Three million pairs with a display are expected to ship this year and twelve million by 2030. Every manufacturer has an answer for text entry, and every answer is: your phone. Where a maker has built something better, it is bonded to that maker’s own glasses and works nowhere else.

This is the part that persuaded us the first two problems were not separate problems. A screen keeps arriving without a keyboard, and the fallback keeps being a rectangle of glass in the other hand.

What we are building

A hand holding an early wand: a small transparent housing with a button, a ribbon cable running out of the back.
An early one, on the bench in Berlin. The ribbon cable is temporary.

A wand. You hold it like a remote, press the button, draw a letter in the air with your arm, and let go. It types — into the television, the laptop, anything that takes a Bluetooth keyboard, with nothing to install at the other end.

No camera. No microphone. It reads the motion of your hand from a sensor that costs a few dollars, and the recognising happens on a chip inside the wand in about 25 milliseconds. Nothing leaves the room, because nothing leaves the wand. Aeroplane mode changes nothing about it.

450 paid strangers have written 52,191 letters in mid-air to teach it. Tested the hard way — leave one person out, train on everyone else — it gets 85% of a stranger’s words right first go, and about 95% once it has learned your hand. If you would like to see what it is actually looking at when it reads a letter, we built a console for that.

Nine things to take away

A card listing nine things in three groups of three: three numbers, three for a cocktail party, and three vibes.
The whole thing on one card, if you would rather have it that way.

Three numbers. 85% of a stranger’s words, first go; 95% once it has learned your hand. 450 paid strangers and 52,191 letters written in the air to teach it. 25 milliseconds to read a letter, on the chip in your hand, with no internet.

Three for a cocktail party. The hardest people to read are the young — under-25s at 55% against 70% for everyone else — because one in five of them writes at thumb scale, a few centimetres of travel per letter, and there is not enough movement in it to read. (Twenty-three people, so a lead rather than a verdict; the full working is here.) Left-handers are nearly three times as likely as right-handers to draw a circle clockwise, and we can see it in the motion. And strangers decided how the wand is held: in the first week of the study the TV-remote grip out-taught the pen grip, and that settled the shape of the thing.

Three vibes. You already know how to use it — you learned at five, with a pencil. Nothing is watching: no camera, no microphone, and that was the first rule, written when the point of the exercise was a game for a small boy. And it does one thing and then it goes quiet — no screen, no feed, nothing to check.

Try it

The recogniser runs in a phone browser. No app, no account, about thirty seconds: write your own name at it. It is the same engine, running on the phone’s motion sensor instead of the wand’s.

Production boards are in build. We plan to show working hardware at CES in January.


A note on the pictures. The woman in the first photograph is not a real person — the image is ours, made with AI, and labelled that way wherever it appears. The wand in the last one is a real wand, photographed on our bench. We would rather tell you which is which than have you wonder.


more from the blog
the wand · blog · press · about · privacy · feed · llms.txt
Input-A (Inputa Inc.) · [email protected]