We Tested Client Sites on Vision Pro So You Don't Have To

The Apple Vision Pro shipped on February 2 at $3,499. The early adopters are filming their airplane seatmates and grocery runs for one another, and the mass-market questions answer themselves for now. This is a first-generation device for developers and enthusiasts, priced and stocked as one. No institutional strategy should be rebuilt around it this year.

That could be the end of it, except for one launch detail that puts this squarely in our jurisdiction. Vision Pro ships with Safari. A real browser on a truly new display and input model. The open web is present in spatial computing’s launch story without anyone’s permission or porting effort. The question ‘what do our sites actually do on this thing?’ can be tested, answered, and answered cheaply. Every device we never planned for is a test of how honestly we built. The face computer says we did fine.

So we spent an afternoon with the device and a representative slice of the institutional web. University program pages, health-system services, commerce templates, content sites. Testing real work on every big new surface has been a habit here since the iPhone post of 2007. The findings are worth your five minutes precisely because most of them are calming and the useful exceptions are specific. Findings first, placement guidance after, pattern note to close.

The reassuring part: standards pay off again

The headline finding, stated with the satisfaction eighteen years of arguing has earned. Responsive, standards-built sites render well in visionOS Safari, untouched. Sites appear as floating windows at user-controlled sizes. The responsive machinery does exactly what it was designed to do. The layouts adapt to the window’s effective viewport the way they adapt to any viewport, because that was always the point. Typography and contrast hold up on the displays. The ordinary institutional site (the program page, the service list, the article) is simply usable right away, on hardware nobody who designed those sites ever imagined.

This is standards-first work paying off again on an unplanned surface, and the mechanism is worth naming because it generalizes. Sites built to the platform’s grain (semantic markup, responsive layout, device-agnostic interaction) inherit new platforms for free. Sites built to specific devices’ assumptions (fixed layouts, hover-only menus, pointer-only behaviors) inherit exactly the debt those assumptions were.

The 2012 responsive commitment has now paid out on phones, tablets, and mobile-first indexing. On foldables nobody optimized for, and a face computer. That is what we mean when we say durable choices keep paying off at moments nobody specifically predicted.

The failures we did find were the predictable class. Hover-revealed navigation: there is no hover in gaze-and-pinch. Tiny tap targets: the WCAG target-size criterion doing double duty again. Interaction patterns that assume a precise pointer. Every one is an accessibility finding, met on a new device. That is the afternoon’s less obvious but important sub-finding. The sites that tested well are the accessible ones. Working on any device and working for any ability turn out to be much the same engineering, as we have argued since touch arrived.

The instructive parts

• WebXR arrives constrained, and the constraint is telling. The immersive-web APIs are present but gated and limited at launch. That signals Apple’s near-term intent, spatial apps through its native frameworks, the web as windows, not worlds. It sets the realistic expectation for any ‘spatial web’ pitch that crosses your desk this year. The platform owner has not opened that door, and strategies premised on it are premised on a roadmap nobody has published.

• Video and 3D content formats show real advantages: well-produced video is impressive at the effective scale the device renders. The panoramic and 3D-model formats the platform favors get real presentation benefits. Media-heavy properties (campus tours, facility showcases, the product-viewing tier) should note the direction. Not as a project this year, but as a format consideration whenever those assets are next produced anyway.

• Gaze-and-pinch is simply another input type. The interaction rules we already follow (generous targets, visible focus, no hover tricks) handle it without new rules. Design for interaction abstractly, and the specific instruments keep taking care of themselves.

Placement guidance, per seventeen years of this genre

The counsel, in the emerging-platform format we have run since the iPhone announcement in 2007. No spatial strategy is required. The installed base is tiny, the use cases beyond early-adopter enthusiasm are unproven, and institutional budgets have better claims on them this year.

One testing afternoon is recommended. The empirical pass above, on your own key templates. The hover-and-target findings are real fixes with benefits on every surface, and firsthand knowledge is cheap protection against both the vendor pitches and the dismissals.

And the watchlist criteria are named now, so future coverage has something to check against. Developer adoption: does a real app world emerge beyond launch-week ports. Price trajectory: the mass-market question is a hardware-cost question on every precedent we hold. And use-case emergence beyond demos: the test that separated the iPhone from the 3D-TV era, did ordinary people find jobs for it. Those three, checked yearly, will tell this readership when spatial computing graduates from watching brief to planning input. Nothing louder than those three should move anyone’s budget meanwhile.

The same test, every device

The method note, because the category is the message. This post’s structure is the same one we ran for the iPhone in 2007 and the iPad in 2010. For the watch, the voice cylinders, and the headsets before this one. Test it yourself. Report honestly. Place it with care. Name the criteria. Decline the hype in both directions. Its record is the reason the structure persists. Enthusiasm pieces and dismissal pieces both aged badly everywhere they were written. The curious-but-unswayed ones aged into useful history.

Vision Pro may be the start of a computing platform, or a superb developer kit for one that arrives later at a different price. We do not need to know which to serve its readers this month. The sites work: credit the standards. The fixes found are worth making anyway: credit the accessibility convergence. The watchlist is set, and the budget is undisturbed.

Stay curious, give nothing away. The pattern has never once failed us, and the afternoon it costs remains the best price in emerging-technology coverage. The device goes back in its remarkable case. The findings are noted, and we will return when the criteria move.