← Module 7/VR/AR: the 2026 reality
RU
Module 7 · The present (2018–26)

VR/AR: the 2026 reality

VR is forever "the future" and forever a niche. The barriers here are not content but physical: presence fights comfort, and the "why" is thin. 2026: the Vision Pro flopped, Quest is the only real market, there has been no hit since Alyx (2020), and the industry is pivoting to AI smart glasses. This is the textbook case of "when a cool technology is not the answer".
~16 min🔬 perception + 💰 market
The gist in 30 seconds
The 2018–22 metaverse boom deflated by 2023–24. The 2026 reality: the Apple Vision Pro flopped (~600k units lifetime, $3500, development being wound down), Meta Quest is the only genuine VR market (~77–84% of shipments, but on a content drought, and Meta renamed "metaverse" to "AI hardware"), there has been no generation-defining title since Half-Life: Alyx (2020), and headset sales are actually falling (−10% to 6.9M/year). Apple and Meta are pivoting from headsets to AI smart glasses (Ray-Ban Meta). The barriers are physical, not content-related: presence demands motion-to-photon <20 ms plus high FOV/refresh (expensive), and comfort fights presence (nausea from vestibular mismatch, vergence-accommodation conflict, weight, isolation) — VR is at war with its own physiology. Plus a thin "why". For a developer in 2026: a niche with a passionate but small audience — it justifies a small dedicated team if the game is built for VR, and never an after-the-fact port. The deeper lesson is that matching a technology to a real need beats chasing the frontier: the hardware analogue of "ML is not always the answer".

The mechanism: why VR is physically hard

Presence — and its brutal budget

The magic of VR is presence: the illusion of "I am there". To fool the brain you need all at once: stereo plus 6DOF head tracking, a wide FOV, a high refresh rate (90+ Hz) and — critically — low motion-to-photon latency: less than ~20 ms between turning your head and the image updating, otherwise the brain catches the mismatch → nausea. The budget is brutal:

tframe= 190 s≈11 ms, tm2p<20 ms

So the whole "sensor → render → display" chain gets ~1–2 frames. Hence the expensive hacks: reprojection / timewarp (nudging a finished frame to match a fresher head pose just before output), foveated rendering, high refresh. Presence is not "a pretty picture", it is hitting a perceptual deadline — miss it and physiology takes revenge.

Comfort versus presence — a built-in conflict

The trouble is that what strengthens presence breaks comfort:

VR cannot maximize presence and comfort at the same time — it is a fundamental trade-off in physiology, not a matter of polish.

The "why" problem

Even with the tech solved, you run into a thin use case. The best one is games built for presence (the physicality of Beat Saber, the spatial interaction of Alyx), but that is a small market. Productivity, social apps and "the metaverse" never took. A $3500 Vision Pro with no killer app is a solution looking for a problem. A technology can be impressive and still answer nobody's real question.

The commercial reality of 2026

The numbers are harsh: the Vision Pro at ~600k lifetime (against hundreds of millions of PS5s/phones), marketing cut, development winding down; Quest holds ~77–84% of shipments but on a content drought, and Meta renamed "metaverse" to "AI hardware"; PSVR2 disappointed; Google and others left VR. There has been no hit since Half-Life: Alyx (2020). Headset sales are falling (−10% to 6.9M in 2024). The main signal is the pivot to smart glasses: light AI/AR wearables (Ray-Ban Meta) instead of immersive headsets; per IDC, the smart-glasses market grew ~211% in 2025. The industry quietly admitted the problem was the brick on your face.

🕹 What to try — and what to notice

VR has to be felt in the body (you cannot explain presence and nausea in text); if you have no headset, look at where the market actually went.

Beat Saber / Half-Life: Alyx presence done right

The benchmarks for "built for VR": Beat Saber is pure physicality (swing your arms — it only works in VR), Alyx (2020) is spatial interaction (grab, turn, aim with your hands) and still the peak of the genre. Neither is possible as a "port" — presence is the gameplay.

🎮 Play: if you have a headset, play Beat Saber and notice that the gameplay lives in your body, not in your fingers. Then try to "walk" with artificial locomotion in any VR game — you will catch a mild nausea; that is why everything uses teleport. Presence and comfort pull in opposite directions right there in your vestibular system.

Apple Vision Pro 2024 · $3500 · a solution with no problem

Technically stunning (4K per eye, eye tracking, passthrough) and commercially a failure: ~600k lifetime, thin software, development winding down. The demo delights for 20 minutes — and does not answer "why would I wear this every day for $3500".

🎮 Watch: if you get a Vision Pro demo, ask yourself honestly after the wow: what killer app would make you wear this daily? The absence of an answer on a $3500 device with the best hardware in the world is the diagnosis: technology ≠ need.

Ray-Ban Meta where the market actually went

While VR headsets stall, light AI smart glasses took off: a camera, an assistant, audio — no immersion and no brick on your face. The smart-glasses market is up 211% (2025). This is Apple's and Meta's bet on the future of wearables: AR/AI, not VR.

🎮 Watch: compare Ray-Ban Meta and the Vision Pro on weight, price and "why am I wearing this". Notice that the winner is not more powerful immersion but less intrusive usefulness. The market picked "always with me and out of the way" over "cuts you off from the world for $3500".

Deep end · perception: latency, vergence-accommodation and simulator sicknessskippable

The motion-to-photon budget

Presence rests on the world staying "nailed" to real space as your head moves. If the tracking→render→light chain takes >~20 ms, the world "swims" behind your head — the brain reads the mismatch as poisoning and switches on nausea. At 90 Hz a frame is 11 ms, so the budget is 1–2 frames for everything. The rescue is asynchronous timewarp / reprojection: render a frame, then right before output nudge it to match the freshest head pose (a cheap 2D correction instead of a new render), pulling perceived latency down. Foveated rendering (sharp only where the eye is looking, via eye tracking) saves budget.

Vergence-accommodation conflict

In reality your eyes converge (vergence) and focus (accommodation) at the same distance. In VR they converge at the object's virtual depth but always focus on a fixed screen a few centimetres away — the brain gets contradictory depth cues → fatigue and discomfort. This is a hardware limit (you need varifocal/light-field displays, which are not yet in mass production), not a software bug.

Why VR specifically makes you sick

Simulator sickness is the inverse of seasickness: there your body moves and the picture is stable; in VR the picture moves and your body does not. Any artificial locomotion (sticks, acceleration) amplifies the mismatch. Hence all of VR locomotion design — teleport, snap turn, peripheral vignetting, "cockpits" as a stable frame: presence sacrificed so you don't throw up. VR design is largely designing against your own physiology.

Deep end · the market: why "there is no VR market, there is a Quest market" and the pivot to glassesskippable

The market structure explains a lot. VR is effectively one manufacturer: Meta Quest holds ~77–84% of shipments, and the rest are either a premium niche (Vision Pro, ~600k lifetime) or gone. "There is no VR market — there is a (small) Quest market", and that one is on a content drought: without hits there is no reason to buy a headset, without an installed base there is no reason to make hits — a chicken-and-egg that has not broken in 10 years.

Why the pivot to smart glasses

The industry reads the signal: the immersive headset lost on three axes — price, comfort/weight, and "why". Smart glasses (Ray-Ban Meta) hit the opposite: cheap, light, out of the way, with a use case of AI assistant/camera/audio rather than immersion. The smart-glasses market is up 211% (2025) against falling VR. This is the classic reversal of "we built it more powerful when it needed to be less intrusive": maximum immersion lost to minimum friction on usefulness.

The lesson for a developer

In 2026 VR is a niche with a passionate, small audience. Building for it is worth it only if the game is built around presence (physicality is the substance of the gameplay), with a small dedicated team. An after-the-fact "let's add a VR mode" port almost never pays back: VR is not a graphics option, it is a different medium with different design.

Analogy
VR is the "flying car" of game technology: forever five years away, dazzling in a demo, but at war with physics (your vestibular system) and economics (price, "why"), which have kept it in a niche for decades. Presence is like a magic trick that works for 20 minutes, and then your neck hurts, you feel queasy, and you remember you are sitting alone with a brick on your face. The pivot to smart glasses is the industry quietly admitting: the problem was the brick, not a shortage of immersion.
Why it matters
VR/AR is the course's best example of impressive technology ≠ needed technology. Decades of hype and billions from Reality Labs did not make VR mainstream, because the barriers are physiological and "why"-shaped, not content-shaped. For an engineer the value is double: knowing the mechanics of presence (latency, vergence, nausea — where VR is genuinely strong), and judging the frontier soberly (when to build for VR, and when it is chasing something shiny). This is the hardware analogue of the course's central thesis: fitting the tool to the task beats fashion, and "ML is not always the answer" = "VR is not always the answer".
🔁 Beyond games — where this transfers
The lesson is matching technology to need, and hype-cycle literacy: the frontier ≠ the solution, and "why" matters more than "wow".

ML / AI (your domain): VR is the hardware twin of "ML is not always the answer", the course's central thesis. A solution looking for a problem, a hype cycle (Gartner: peak of inflated expectations → trough → plateau) and the "metaverse hangover" are exactly the pattern of AI hype waves and the over-promise/AI-winter risk (see classical vs ML): the temptation to bolt an LLM onto a place where regex or a classical method would do is "a VR mode as an after-the-fact port". The perceptual deadline of presence (<20 ms motion-to-photon) ⇄ latency budgets in interactive AI (the 3 s threshold for an LLM NPC): past the threshold the illusion collapses. And the VR→AI-smart-glasses pivot is a signal of where the frontier actually went (on-device assistant, unobtrusive usefulness), visible only if you read the market rather than the hype. The professional skill is hype-cycle literacy: don't bet a studio or a project on the peak of inflated expectations.

Product / strategy: "a technology looking for a problem" versus jobs-to-be-done; the platform chicken-and-egg; unobtrusive usefulness beating maximum power (smart glasses vs headset).

Hardware / HCI: perceptual thresholds (latency, resolution, vergence) as hard limits; designing against the user's physiology.

Principle: impressive ≠ needed. Ask "what real job does this do and at what cost in friction", not "how cool is this on the frontier"; read the market, not the hype.

🔧 Run it and poke at it — on your home machine
What to play is above (🕹). Here — hands on the engineering and on judging the frontier.
🔧 Tinker (if you have a headset) ~40 min, Quest / SteamVR
In any VR game, find the locomotion and comfort settings: switch teleport↔smooth movement and vignetting on/off — feel the presence↔nausea trade-off yourself. Turn on the latency/frame-rate readout (if there is one) and move your head sharply — catch the moment the world "swims". That is the motion-to-photon budget live.
🧪 Test (judging the frontier) ~15 min, no headset
Take any "hot" technology (VR, or a fresh AI frontier) and run it through three questions: (1) what real job does it do better than what exists? (2) what is the cost in friction (money, comfort, latency, learning)? (3) where is it on the hype cycle? Then: for an imaginary game — is a VR version worth making, and what in it would have to be "built for presence" so it is not a port?
Checklist: felt the presence↔comfort trade-off (teleport/vignette); caught the world "swimming" on a sharp movement; ran a technology through job/friction/hype cycle; decided VR yes/no for a hypothetical game with a rationale.
Connections
foundation
The 3D pipeline — VR is 6DOF stereo rendering under a hard latency deadline: the same pipeline ×2 eyes plus timewarp.
contrast
Classical vs ML — the same thesis, "a cool tool ≠ the needed tool": VR is to presence what ML is to a task where classical methods are enough.
adjacent
LLM NPCs — the perceptual threshold: presence collapses past 20 ms, the illusion of a living NPC past 3 s. Latency deadlines for interactivity.
Questions worth asking
Why has VR failed to go mainstream for decades — surely it is not just content?
Content is a consequence, not a cause. Three fundamental barriers: physiology (presence and comfort pull in opposite directions — nausea, vergence-accommodation, weight/isolation; polish does not cure this), the "why" (a thin killer use case outside niche games) and price/chicken-and-egg (expensive hardware → few headsets → little reason to make hits → no reason to buy a headset). The content drought is a symptom of that tangle, not its root. That is why "even more powerful hardware" (Vision Pro) did not help, while the turn to unobtrusive smart glasses did — it removes price, comfort and "why" at once.
Why 20 ms specifically — what physically happens at higher latency?
Presence rests on the virtual world staying "nailed" to space as you turn your head. The visual system is extremely sensitive to a mismatch between the vestibular signal (head rotation) and the image update: past ~20 ms of delay the world visibly "swims"/lags, the brain reads it as a sensory conflict (as in poisoning) and switches on nausea. At 90 Hz one frame is 11 ms, so the whole sensor→render→display path gets 1–2 frames. That is why VR engines resort to reprojection/timewarp (nudging a finished frame to a fresh pose right before output) — anything to keep perceived latency under the threshold. 20 ms is the perceptual deadline past which the illusion physically collapses.
Is the Vision Pro's failure about the price ($3500) or about VR's fundamental problems?
Price is an amplifier, not the root. Even a free Vision Pro would hit the same barriers: heavy, isolating, vergence-accommodation fatigue, and above all — no killer app worth wearing it daily for. Apple delivered the best hardware in the world (4K per eye, eye tracking) and proved that the problem is not the quality of the immersion: the device is technically magnificent and commercially dead. That is the diagnosis — VR is blocked not by "not made well enough" but by physiology and the absence of a need. $3500 simply made the failure fast and obvious.
Why is the industry pivoting to smart glasses if VR immersion is "cooler"?
Because "cooler" turned out not to be what was needed. An immersive headset maximizes presence at the cost of price, weight, isolation and "why". Smart glasses (Ray-Ban Meta) go the opposite way: minimal immersion, but cheap, light, always with you and out of the way — and the use case (AI assistant, camera, audio) is real and daily. The market picked unobtrusive usefulness over maximum power (+211% smart glasses against −10% VR). This is a recurring hardware lesson: what wins is not the peak of capability but the best ratio of usefulness to friction. The same logic by which the phone won and Google Glass and 3D TV did not.
When SHOULD a developer build for VR after all?
When presence is the gameplay rather than a decoration: the physicality of Beat Saber (you cannot swing your arms outside VR), the spatial interaction of Alyx (grabbing and aiming with your hands), simulators (cockpit, training), horror (bodily fear). The marker of "built for VR": take the headset away and the game stops existing, rather than "loses a bit of wow". Then a small dedicated team aimed at a niche but passionate and paying audience is justified. What almost never pays is the after-the-fact port, "let's add a VR mode to a normal game": VR is not a graphics option but a different medium with different locomotion, UI and scale design; such a port usually does not pay back and suffers on comfort.
Further reading / watching