Vision Pro Was Supposed to Kill the Window. Two Years Later, the Window Won.

Forty-five thousand. That's how many Vision Pro units IDC projected Apple would ship in the 2025 holiday quarter — historically Apple's biggest sales window of the year, for a device that was supposed to replace the laptop, the TV, and eventually the phone. Fintool reported that production had already halted by January 2026, with total 2025 shipments landing somewhere around 80,000 to 90,000 units. Compare that to the roughly 390,000 units Apple's manufacturing partner Luxshare shipped during the 2024 launch window, and you're looking at a product that lost most of its own demand in eighteen months.

Everyone remembers the sales story by now. Fewer people are willing to say the design story out loud, because a lot of us wrote it. In early 2024, the design community produced a wall of confident predictions: windows would float free of the desktop metaphor, pinch gestures would replace clicks entirely, infinite spatial canvases would replace the flat page. None of it happened at the scale anyone promised. The apps that survived on visionOS didn't lean harder into spatiality — they backed away from it. And that's not a footnote to the Vision Pro's commercial failure. It's the actual design lesson, and it's more useful than "the headset flopped."

What the 2024 Spatial-UI Predictions Actually Said

Go back and read the design blogs from launch week and the confidence is almost funny in hindsight. Vision Pro coverage from that period described a shift from "pixels on a page" to "objects in a room," with 2025 pitched as the year UX would fully embrace spatial-first thinking and stop treating apps as flat rectangles. The pinch gesture got sold as a frictionless replacement for the click — no physical effort, hands resting in your lap, eye tracking doing the targeting work for you. Voice, gaze, and gesture were going to eliminate the "physical barriers" between people and their content, per the same wave of coverage.

Apple's own marketing fed this directly. VisionOS windows weren't windows, they were "planes in space" you could push into a room and walk around. The pitch wasn't an upgrade to existing app patterns. It was a replacement for them.

The Sales Collapse Nobody in Design Wanted to Read As a Design Signal

The commercial numbers get treated as a hardware story — too heavy, too expensive, no killer app. All true. Morgan Stanley analyst Erik Woodring told the Financial Times, as reported by PYMNTS, that "the cost, form factor and the lack of VisionOS native apps are the reasons why the Vision Pro never sold broadly." Notice the third item on that list. It's not a chip problem or a battery problem. It's a content problem, and content problems are design problems wearing a business-page disguise.

Counterpoint Research found the broader VR headset market fell 14% year over year, with Meta's Quest line holding roughly 80% of unit sales — meaning Vision Pro wasn't just underperforming its own targets, it was losing share inside an already-shrinking category. Apple slashed digital ad spend for the device by more than 95% in the US and UK, per the same reporting. You don't pull marketing that hard on a product you still believe needs evangelizing. You pull it on a product where the app store isn't giving people a reason to stay.

And developers noticed the emptiness before the sales numbers confirmed it. Coverage from early 2025 described developers as unsold on the platform as available apps actually declined post-launch rather than grew — the opposite of the curve every successful platform needs. One independent developer, writing publicly under the FlineDev name, described walking away from visionOS after hitting a wall of missing APIs that made it impossible to build the immersive AR experiences the platform had promised in the first place. That's the tell. It wasn't that gesture-first design was hard to build. It's that the infrastructure to build the launch-era vision never fully showed up, and the people closest to the platform stopped waiting for it.

If you want a broader read on how AI and platform teams keep shipping the announcement before the substance, Codexical's piece on deployment theatre covers the same pattern from a different angle — the gap between what gets marketed and what gets delivered.

The Apps That Actually Worked Went Flatter, Not More Spatial

Here's the part that should have been the headline all along. Apple's own Human Interface Guidelines for visionOS, updated well past launch, say directly that "not every moment in a visionOS app needs to be fully immersive" and that a windowed, UI-centric experience is often the right call — because windows are easy to move, organize, and control in ways pure spatial environments aren't. That's Apple's own design team walking back the launch rhetoric in its own documentation.

VisionOS 26 shipped spatial widgets as an incremental feature, not a category overhaul — small dashboard-style panels that add a bit of depth to an otherwise familiar layout, according to AppleInsider's coverage. Ornaments, the floating toolbars that sit just outside a visionOS window, exist specifically so the main content area can stay a flat, familiar panel while only the peripheral chrome gets the spatial treatment. That's not "objects in a room." That's a 2D app with a garnish.

None of this is designers failing to execute the 2024 vision. It's designers finding out, the expensive way, that people don't actually want to relearn interaction from zero. A calendar is legible because it's a grid. A text field is legible because it's a rectangle you tap into. Strip those conventions away in favor of "natural" gesture and gaze interaction and you don't get intuition — you get a device people describe, accurately, as tiring to use for more than twenty minutes.

The Turn: Spatial Interfaces Win By Disappearing

So the honest read on 2024's design predictions isn't "spatial computing is dead." Hardware categories survive bad first products constantly — the first tablets flopped too. The honest read is that the predictions had the mechanism backwards. Spatial-native design rhetoric assumed users wanted to feel the spatiality: windows drifting, gestures replacing familiar inputs, environments announcing themselves as new. What actually shipped and stuck did the opposite. It kept the flat panel as the unit of legibility and used spatial depth only where it reduced friction instead of adding spectacle — an ornament here, a widget with a little parallax there, immersive mode reserved for the handful of moments (a movie, a game) where full immersion is the point rather than the gimmick.

This isn't a new discovery in interface design. It's the same correction the industry made with skeuomorphism a decade earlier — physical metaphors are useful exactly as far as they do cognitive work, and worthless the moment they become decoration. Codexical covered that cycle in detail: flat design won in 2013 because ornamentation without function is friction, then partially reversed a decade later because some of that ornamentation had been doing real affordance work all along. Spatial computing is running the identical arc on a compressed timeline. The apps that got the memo stopped trying to prove the platform was different and started asking, panel by panel, whether depth was earning its keep.

The best spatial app you'll use this year won't look like a spatial app. It'll look like the software you already know, with just enough depth added that you never notice the difference — which was always the actual bar, and never the one anyone put on a keynote slide.