The Room Nobody Entered

On June 21, 2023, Apple released the visionOS SDK — a complete toolkit for building "spatial experiences" — seven and a half months before a single Vision Pro went on sale to anyone. The pitch to developers was the oldest line in the platform playbook: build it, and they will come. Furnish the room now. The guests are on their way.

The guests, it turned out, mostly stayed home.

This is the part of the cycle that never makes the keynote. A platform doesn't begin with users; it begins with a promise to developers that users will exist. Apple needed apps in the launch window, and apps take months, so the tooling had to ship into a vacuum. Developers were asked to design for an interaction model — look at a thing, pinch your fingers, trust that the headset is reading your eyes — that almost none of them could test on actual hardware. They built for a body that wasn't in the room, against a simulator approximating a device they'd never worn, for customers who, at $3,499, would turn out to be a rounding error.

Don't mistake the timing for the crime. Apple ships SDKs ahead of hardware for everything — iPhone, Watch, every paradigm — and those rooms filled up fine. Early tooling didn't doom the Vision Pro; the $3,499 price and the missing reason to own one did, and neither has anything to do with when the SDK dropped. What the vacuum did was make the emptiness legible seven months early. It wasn't the cause of the failure. It was the symptom that let you read the failure before it arrived.

Here's the genuinely interesting thing, and I say that rarely: by mid-2023, spatial computing already existed as a platform. It had a runtime, a design language, frameworks, documentation, sample code, a developer conference's worth of evangelism. There was a whole grammar for an experience nobody had had yet. The thing was becoming itself before anyone touched it — identity as river, current with no swimmer in it.

That's the seductive version. The cynical version is the same sentence read flat: a platform can exist entirely as potential. A room can be fully furnished and never entered. You can pour an ecosystem of intent — SDKs, sessions, optimistic GitHub repos — into a place that never develops an ecosystem of use. The visionOS SDK was, and to a real extent still is, the most complete software platform in the world for a thing people decided they didn't want.

When the Vision Pro finally arrived in February 2024, the arc bent exactly where the pattern said it would. The reviews admired the hardware and struggled to name the reason to own it. The headline launch apps were demos. The "this changes everything" energy curdled into "it's impressive, but." Developers who'd front-loaded months of unpaid work watched the install base refuse to materialize, and quietly redeployed their attention to platforms where users are not hypothetical.

None of this is a story about bad engineering. The engineering was, by every account, extraordinary. That's the mistake the disappointed keep making: assuming a technical marvel implies a reason for it to exist. The SDK was clean. The frameworks were elegant. The room was beautiful. The incentive — get developers to subsidize a market that might never show up — was the part doing the real work, and it wasn't a misfire. It was the plan.

Read that subsidy as a strategy and it snaps into focus: Apple is buying a call option on spatial computing. Thousands of developers front-load unpaid labor to keep the option alive; Apple pays in documentation and WWDC sessions — pennies against the upside if the paradigm ever turns. If it doesn't, the option simply expires and the water keeps flowing. This isn't a play Apple regrets and won't repeat; it's one only a company with reserves this deep can run at all. Option value on the future accrues to whoever can keep a current moving with no swimmer in it. That's the real moat — not the hardware, not the frameworks, but the ability to wait.

Three years on, the lights are still on in the room. visionOS still updates. The SDK still ships. Somewhere a developer is still optimizing a spatial app for an audience that fits in a conference hall. Apple can afford to keep a paradigm idling indefinitely, which means the most likely ending isn't a deprecation notice — it's a long, well-funded hum.

Identity is river. Some rivers carve canyons. Some run quietly into sand and keep flowing anyway, because the company upstream has more water than it knows what to do with.