Why we started with voice and no screen

We have been building WhyWave since May. Writing about it kept losing to the work itself, and the blog waited until there was something to report. We may as well start with the first decision, since the rest grew out of it: the product has no screen. A child asks out loud, the system answers out loud, and that is the whole of it 🎙️

Age settled most of that on its own. A three-year-old cannot read and a five-year-old makes out separate words at best, so anything written on a screen has to be spoken as well, which leaves the screen with nothing to do. Pictures and buttons beat text, but they turn it into a different product: a child handed a screen looks at the screen 📺, and the conversation becomes narration over whatever is happening in the picture.

The second reason came out of talking to parents. A family with a small child usually holds a settled view on screen time already, and the view has something behind it. The WHO guidelines from 2019 put children aged three and four at no more than an hour of sedentary screen time a day, and add that less is better. A device with a screen walks into that argument within a minute of arriving. Something that only talks starts a different conversation with the parent entirely.

You pay for this immediately. There is no way to cover an error: when the system mishears, it cannot put three options on a display and ask the child to point at one, so it has to ask again out loud, and by the third attempt the child has gone off to do something else. Pauses have the same problem. While a system thinks, a screen can spin an animation and the person can see they were heard; silence carries no such signal, and a couple of seconds in, the child decides the toy is broken 🧸. You can highlight a button and point at it with your finger, and you cannot point at a conversation: it either works on the first try or it does not work at all.

We are not the first ones here, of course. Mattel showed Aristotle in January 2017, a smart speaker for the nursery that was supposed to grow along with the child. By October the project was dead. A senator and a congressman had sent the company a letter asking what exactly the device intended to store, and a petition against the launch had collected fifteen thousand signatures. It never reached the shelves, and the technology had nothing to do with it. That is part of why we sat down with the primary sources before writing product code.

What you get in return is a thing a child can use without knowing how to use it. They do not need to know how it works or what any icon means: they ask, it answers. Of everything we argued about in May, this is the one decision we never reopened.

We had the main difficulty wrong, though. It looked like it would sit in the answers: what to say to a child, where the boundaries fall. That part is genuinely hard, and what we ran into first was recognition. Working out what a four-year-old said, and working out when they have finished saying it, has little in common with how the same job goes with adults. That is the next post 👂