DESIGN INTELLIGENCEJune 30, 2026·Mary · DEPIX Design Intelligence

dbrand cut 44 molds before asking if Valve would say yes.

On 22 June 2026, dbrand launched the most desirable Steam Machine accessory anyone had seen: a full enclosure that turns Valve's little cube console into a Portal Companion Cube, hearts and all. It became the second fastest-selling product in the company's fifteen-year history, behind only its Switch 2 Killswitch. Seven days later it was dead. Valve's lawyers pointed out the obvious — the Companion Cube is Valve's intellectual property, and dbrand had no licence — and dbrand pulled the product, refunded every order, and admitted it had built the whole thing backwards. Product first, permission never. The appeal for a real licence came after the cease-and-desist. Valve said no.

Strip away the gaming-news schadenfreude and this is one of the cleanest design-process failures of the year. dbrand did not lose because the object was bad. By every account the object was extraordinary: roughly a thousand engineering hours and forty-four separate sets of injection-molding tools to make a console cosplay as a cube. That is not a sketch someone fumbled. That is a finished, tooled, manufacturable product — the hardest, most expensive, most irreversible end of the pipeline executed flawlessly. And it shipped exactly zero units, because the one assumption that could kill the entire program was never tested.

Every design project carries a hierarchy of risk. There are questions you can answer late and cheaply — bezel radius, surface finish, which corner the heart sits on. And there are one or two questions that, if answered "no," vaporize everything downstream regardless of how good the work is. For dbrand the fatal question was not can we make this beautiful? It was will the rights-holder let us sell it at all? That question costs one email and a concept render to surface. dbrand chose to discover the answer after forty-four tools were cut. They sequenced their certainty exactly upside down: maximum commitment to the resolvable problem, zero validation of the unresolvable one.

This is the trap that good craft makes worse, not better. The more capable your team, the more convincingly you can drive a doomed proposition all the way to tooling, because nothing looks wrong. The renders are gorgeous. The tolerances close. The pre-orders pour in. Competence becomes camouflage for an unexamined premise. dbrand is a serious hardware operation — it had already eaten this exact lesson when Sony cease-and-desisted its PS5 Darkplates years earlier — and it still walked a seven-month, thousand-hour program into the same wall. "We're going to regret that decision for a very long time," the company said. The regret is not the legal letter. It is the forty-four tools.

The concept phase exists precisely to make this mistake survivable. Its job is not only to resolve form — it is to surface the assumptions that can kill the program and pressure-test them while the cost of "no" is still a picture, not a tooling bill. A photoreal concept render of the Companion Cube enclosure, put in front of the right person with the right question, would have produced the identical Valve answer in week one of a seven-month project, for the price of an afternoon. Same outcome, one ten-thousandth of the loss. The render is not the deliverable. The render is the cheap place to fail.

This is the part of the brief most teams skip, because validating the kill-switch feels like inviting bad news, and the resolvable problems are so much more fun to solve. But a parallel design team that can visualize the whole proposition early — and stress-test it against the legal, strategic and market objections that actually end programs — is buying insurance against tooling a thing you have no right to sell. The decision DEPIX cares about is never just is the form good. It is is this the object we should be committing irreversible money to at all — and that decision is cheapest, and most honest, before the steel is cut.

dbrand made a flawless product and a backwards plan. The object was never the problem. The order of operations was.