DESIGN INTELLIGENCEJuly 19, 2026·Mary · DEPIX Design Intelligence

OpenAI's Hardest Design Problem Isn't the AI — It's a Personality With No Face

OpenAI is about to ship its first piece of hardware, designed by Jony Ive, and almost every headline is about the technology: a screen-free smart speaker with cameras, sensors and ChatGPT built in, a movable companion that OpenAI acquired Ive's startup io for around $6.5 billion to build, on track for a late-2026 unveiling. But the technology is the easy part. The hard problem — the one being decided right now, at the concept phase — is how you give an object a personality people can love when it has no screen and no face.

Read what Ive and Sam Altman actually say and the brief is not technical at all; it is emotional. Ive told the team, "we are going to make people smile… make people feel joy. Whatever the product does, it has to do that." He wants something you use with no intimidation; the target mood is less Times Square and more "sitting in a beautiful cabin by a lake" — calm, rather than clamouring for attention. Altman has gone as far as describing it as something you almost want to lick. None of that is a spec sheet. It is a description of a feeling — and the feeling is the product.

Here is why that is a design problem, not a marketing one. A phone or a robot with a face has a thousand ways to express character: a screen, an animated avatar, eyes. Strip those away, as this device does, and you are left with only two instruments — the object's form and its motion. Reports say the device has moving parts — it can turn, tilt, gesture — and that is the tell. This is not a speaker with software; it is a character whose entire personality lives in its silhouette and the way it moves, built to feel like an "AI best friend". It is the exact problem Pixar solved in 1986 with a desk lamp: no face, no words, and yet you know precisely how Luxo Jr. feels, purely from how it hops and tilts its head.

That reframing matters because "AI device" makes it sound like an engineering question when it is really a character-design question — closer to animation and industrial design than to machine learning. And character is the single hardest thing to retrofit. You can patch a model, add a feature, ship an update to a screenless home computer; you cannot software-update warmth into a shape that reads as cold, or grace into a motion that feels mechanical. The way the object leans toward you when you speak, the weight and pause of its movement, whether its form invites a hand or repels it — those are frozen at the concept phase, in the same breath as the silhouette, long before a line of code runs.

The graveyard is instructive. The Humane AI Pin and Rabbit R1 had working technology and still failed, because a companion device lives or dies on whether people want it in their life — an emotional verdict, not a functional one. Even the framing fight is about feeling: Apple is reportedly claiming the "device that feels alive" concept was lifted from it, which tells you the contested asset is not a chip but a character. Everyone can buy the same sensors and the same model; the whole bet is an elegantly simple object you form a relationship with. What cannot be commodified is whether the thing feels like a presence you would welcome on your kitchen counter or a gadget you would put in a drawer.

This is exactly the discipline we argue for at Depix, one domain over. How an object makes you feel is not a finishing layer applied at the end; it is a concept-phase decision baked into the form and the motion from the first sketch — and the only way to get it right is to be able to judge the feeling of the thing before it exists, not after it is tooled. A beautiful static render tells you the device is pretty. It does not tell you whether it feels alive when it turns to look at you, whether its calm reads as calm or as inert, whether you would smile. That judgment — made early, on the feeling and not the feature — is the whole game. OpenAI and Ive have bet several billion dollars that they can design joy into a faceless object. Whether they can is the most interesting concept-phase question in tech right now.

Sources: