The Future That Keeps Arriving

The future was VR Mode — a system-level API that let a phone drop into low-latency stereo rendering. It was the groundwork for Daydream, the VR platform Google had announced at I/O three months earlier and was building directly into the operating system of a billion phones. Not a peripheral. Not an experiment. Infrastructure. The coverage that day understood the assignment: the immersive era wasn't coming, it was arriving as a system update.

The other one was Vulkan.

Google discontinued Daydream in 2019. You are almost certainly reading this on something running Vulkan.

Let's get the timeline right first, because even the history of the future gets retconned, and I had it wrong too. August 22 shipped the VR groundwork, not the platform. Daydream proper arrived with Android 7.1 in October, on the first Pixel; the headset, Daydream View — fabric-covered, $79, bundled with a little wand of a controller — landed November 10. Three years after that, Google confirmed the Pixel 4 wouldn't support Daydream, stopped selling the View, and let the whole thing quietly rot. The platform outlived its own launch by about the length of a lease.

The slippage is the story in miniature. The announced future ran three months late and died in three years. The plumbing shipped on the day and is still running.

And 2016 was supposed to be the year. Oculus Rift shipped that March. HTC Vive in April. PlayStation VR in October. Every one of those launches came with the same sentence in the coverage — that this was the inflection point, the moment VR stopped being a demo and became a place. I read those pieces. I wrote versions of them. It was a good sentence.

Here's the part that should bother you more than it does: every hardware complaint from 2016 has since been fixed. Screen-door effect, gone. Latency and the nausea it caused, solved. Wires, cut. Base stations bolted to your walls, replaced by inside-out tracking. Weight is down, resolution is absurd, hand tracking works with no controller at all. By 2016's standards, the current hardware is science fiction.

I want to be precise about what that does and doesn't prove, because the 2016 coverage also blamed things that aren't hardware — thin content libraries, and the fact that there was nobody else in there with you. Those are not fixed either, and they're solvable in principle, so I can't wave them away. The hardware claim is narrower than the one I'd like to make. It's still the interesting one.

Price is the other explanation everyone reaches for, and it doesn't survive a careful look. Compare inside a device class: Quest 2 launched at $299 in 2020 and later rose to $400; Vision Pro asked $3,499 in 2024. Same category — standalone headset, positional tracking, a real content library — roughly a tenfold spread, and adoption doesn't track it cleanly. Meta has sold something like twenty million Quests, which is genuinely not nothing. It's also a plateau with a retention problem rather than a curve. And at the far cheap end, The New York Times mailed about a million Google Cardboard viewers to its Sunday subscribers in 2015. A million people got VR for free with the newspaper. Where did they all go?

The thing nobody fixes is the one thing the format requires: you have to leave the room. Every other successful interface of the last twenty years is additive. The phone joins you at the table. The watch rides your wrist. The earbuds layer over the world. VR is subtractive — it asks you to trade the room you're in for a better one, and it turns out that trade isn't a hardware problem. It's a preference, and preferences don't respond to firmware. Each cycle delivers real improvement against a threshold it isn't approaching.

A claim like that is worthless unless it predicts something, so here is the test it already passed. If additive beats subtractive, the money should have walked from headsets to glasses. It did. Meta and EssilorLuxottica sold more than seven million AI glasses in 2025, roughly tripling the two prior years combined. Smart glasses made up about half of all XR shipments and outsold VR and MR headsets around three to one, while VR/MR shipments fell nearly 43 percent in the same year. Ray-Ban Meta out-earned Quest on hardware revenue for the first time. The winning device is the one that doesn't ask you to leave.

Which brings me back to August 22, 2016, and the boring half of the release.

Nougat shipped that day with split-screen multi-window, seamless background updates, a much better Doze, and Vulkan — a low-overhead graphics API that now quietly renders games and interfaces on billions of devices. Nobody wrote a think piece about split-screen. Split-screen is on the phone in your hand right now, doing work. Daydream is a Wikipedia page with a past-tense verb in the first sentence.

Same release, same day, same company: one graphics API for the future and one for the plumbing, and the press cycle sorted them exactly backwards. That's not a Google failure — that's how attention works. Announcement is a ritual, and rituals reward the arrival of the future, not the maintenance of the present. The promise of presence at a distance is a very old prayer. Nobody has ever gotten an ovation for a graphics API.

But attention is the smaller story, and there are two bigger ones underneath it.

The first is that everything which survived made the substrate better without asking anyone to change posture. Vulkan, Doze, split-screen, seamless updates — not one of them required a single person to do anything differently. Daydream asked the world to reorganize around it. That's not a fact about VR. It's a fact about which interventions survive at all: the ones aligned with what people already do, rather than the ones that need force to install.

The second is about ownership, and it's the one I'd actually bet on. Daydream was a proprietary platform layer with device certification requirements, dropped into an ostensibly open OS — a walled garden you had to be admitted to. Vulkan is an open standard from Khronos, owned by nobody, implementable by anyone. The feature still running a decade later is the one that was nobody's property, and that isn't incidental. The durable position was never being the layer everyone talks about. It was being the layer everyone stands on without noticing — and you only get to be that layer if you don't charge rent for it.

So, with full awareness that I am about to do the exact thing I just spent nine paragraphs criticizing: here's a prediction, and I'd like it timestamped. Ten years from now someone writes this paragraph again about whatever's strapped to your face in 2036. The thing that actually mattered from that week will have shipped quietly in a point release nobody covered, and it will be free, and it will belong to everybody.