Where It Lands

There is a Greek word for a thing that is both the cure and the poison, and nobody has been able to translate it for two and a half thousand years.

The word is pharmakon, and the reason it resists translation is that it refuses to hold still. Plato uses it for writing. Writing, says the god who invents it, is a remedy for memory. Writing, says the king who refuses it, will destroy memory, since men who can look a thing up will not trouble themselves to hold it. Both are correct. The word contains both and will not choose, and translators have been quietly choosing on its behalf ever since, which is its own small scandal.

Here is why this matters to anyone standing in a room with fifteen headsets and a folding table.

A pharmakon is not a substance. It is a dose. Its character is not a property you can find by examining it more closely; it is a property of the scene in which it is administered — who hands it to you, how much, for how long, next to whom, and what happens in the twenty minutes after. Nothing about the molecule tells you whether it heals. You have to look at the room.

Which is to say: the question of whether virtual reality is any good is not a question about virtual reality. It never was. It is a question about distribution, and distribution has been treated for a decade as the boring part — logistics, the loading dock, the thing you do after the art is finished.

The art was never finished. The loading dock is where it gets made.

Somebody will tell you the medium is dead, and they will be holding a real number when they say it.

Eighty-eight billion dollars, give or take, has gone into Reality Labs since late 2020. Nineteen point two billion of it in 2025 alone. In January of 2026 the company cut roughly fifteen hundred people from the division, shuttered studios with names like Twisted Pixel and Sanzaru, retired the office-meetings app, and moved its flagship social world into what the trade press calls maintenance mode, which is the industrial term for a coma with the lights left on. On the earnings call the word metaverse went quiet and the word glasses got loud. Glasses sales tripled.

Note the grammar of the obituary. Not the medium died — an event, with a date. The medium is dead: present tense, a standing condition, the tense we reserve for things we would rather not have to actually bury. Which is convenient, because the corpse keeps making noise. Festivals are full. Works are winning prizes. People in New Haven and Torrington and the Tenderloin are putting these things on other people's heads on a Thursday evening and the conversation afterwards runs long.

So the honest description is not death. It is hauntology — not a medium that stopped, but a future that stopped, while the artifacts it produced went on working perfectly well and nobody could quite explain what they were for anymore.

The bridge everybody keeps proposing between French cultural funding and American immersive art already exists. It was built around 2020. One end walked away.

Consider two works that have been shown together, in San Francisco, in 2026, to ordinary people who then talked to each other for an hour.

Goliath: Playing With Reality — twenty-five minutes, animated, narrated by Tilda Swinton, about a man diagnosed with schizophrenia who finds company in multiplayer games. It won the Grand Jury Prize at Venice in 2021. Its funders, and read this list slowly: Oculus VR for Good, the British Film Institute, Creative XR, Digital Catapult, Arts Council England, the CNC, and StoryFutures, co-produced with a Paris company.

That middle acronym is the hinge of this whole document, so: the CNC is France's national agency for cinema and moving images, and it is funded not by the state's general budget but by a slice of every cinema ticket and streaming subscription sold in the country. The industry taxes itself and the proceeds go back into making things. Which means French support for a medium does not have to wait for the medium to be profitable — a structural fact that looks, from the outside, indistinguishable from taste.

We Live Here — Rose Troche, premiered at Venice in 2020, made with Invisible People. It puts you in Rockey's tent after a sweep has driven her out of it, and lets you handle her belongings until they give up her life. Also an Oculus VR for Good project.

So: the Franco-American immersive bridge is not a proposal anyone needs to write. It was poured in concrete around 2020, half of it with French state money and half of it with Menlo Park philanthropy, and then one abutment quietly withdrew and the works were left standing on it in mid-air with nowhere to go.

They are not unavailable. That is the strange part. They are free, they are finished, they are award-winning, and they are unseen — a decade of grant-funded, artistically serious immersive work that exists in a state of permanent near-availability, like a library with no address.

Every account of this technology is an account of addition. More presence, more immersion, more world. Phenomenologically it begins with subtraction.

Put one on. The room goes. The people beside you go. Market Street goes. Your peripheral vision, your sense of scale, your ordinary way of moving — all revised or removed by somebody else's software. Only then does the other world arrive, and it arrives in the space made by what left.

The device is a mask. Not metaphorically; structurally. It occludes, it frames, and only in that order does it reveal. Which suggests a proposition with a much wider blast radius than one industry:

Every perspective reveals by occluding. The headset is simply the only one you can hand to someone else.

The temptation here is to run the old routine — mask off, true self underneath, applause. Resist it. There is no unmasked condition. Attention filters. Language filters. Expertise filters, and rather ruthlessly. Trauma filters. Institutions filter, that being roughly what an institution is. The problem was never that we wear masks. The problem is forgetting that the mask is a mask, at which point it stops being a perspective and becomes simply the way things are.

Taking one off does not deliver you into unmediated reality. It restores the possibility of changing masks. Which is a smaller claim, and survivable.

And now the part with a clock on it. A headset is a mask with a removal event built into it — a physical gesture, visible to everyone in the room, that ends the state. Smart glasses are a mask engineered never to come off. The industry did not abandon mediated perception in 2026. It abandoned the version of mediated perception that has an ending, in favour of one that has a battery life.

The word we built the industry on means the opposite of what we used it to mean.

In 2015 a filmmaker stood on a TED stage and called this the ultimate empathy machine, and an entire funding category grew out of the sentence. The phrase itself was secondhand — Roger Ebert had used it about cinema — and the concept underneath it was thirdhand and worse.

Einfühlung, the German term that gave us the English word, was coined by aestheticians describing what happens when you look at a column and feel it straining upward. It named the viewer projecting their own feeling into an object. Not access to another mind. The precise opposite: the reliable human habit of finding yourself in things that do not contain you.

So the founding claim of the field is a mistranslation twice over — once out of German, once out of film — and the thing it names, on the evidence, is the exact hazard the field should have been guarding against.

The evidence, since somebody should cite it: a pre-registered meta-analysis of 43 studies and 5,644 participants found that VR reliably raises emotional empathy and does essentially nothing for cognitive empathy — it moves your feelings and leaves your model of the other person where it was. Against controls it beat nothing, and beat video, and beat being told to imagine. It only marginally beat reading about someone. And against witnessing another person in actual life, it scored d = −0.09, which in plain speech means: slightly worse than looking at somebody.

The correct response to this is not despair, and it is certainly not to run the study again until it behaves. The correct response is to notice that we measured the wrong thing. Nobody at these gatherings is claiming to have acquired another person's inner life. What they are doing is staying in the room afterwards, which no instrument in that literature was pointed at.

A sentence Derrida spent thirty years being misquoted about turns out to be a practical instruction for people who book venues.

Il n'y a pas de hors-texte. Rendered into English as there is no outside-text, then flattened into everything is text, then weaponised by people who had not read it into nothing is real, which is not what it says and never was.

He eventually got exasperated enough to gloss it himself, in the afterword to Limited Inc: the phrase, he wrote, means nothing other than there is nothing outside context.

There is a joke buried in the French that survives translation badly. Hors-texte is a printer's term. It is the unpaginated matter bound into a book — the plates, the frontispiece, the material that sits outside the page numbering while being physically, undeniably, inside the volume.

Apply that to a screening. The venue. The eleven people who came because it was snowing and they had nowhere else to be. The person who put the headset on your head with their own hands because you couldn't work the strap. The eight dollars instead of ten because you live in the neighbourhood. The argument in the doorway on the way out.

These are not the packaging around the work. There is no packaging. They are bound into the volume.

Which produces, from a piece of French philosophy that has been accused of leading nowhere in particular, an unusually concrete conclusion: whoever composes the context is an author. Not a host. Not a service provider. Not the logistics. The person who chooses the room, the group, the sequence and the moment of removal is doing authorship, and the only reason it has never been funded as authorship is that we did not have a word for it.

A caution, and it is aimed at this document. None of the people doing this work were waiting for a philosopher to authorise it. The practice is years ahead of the theory and always was — the rooms were being composed, carefully, by people who could already tell you exactly why the bookshop worked and the museum atrium didn't, long before anyone reached for a French footnote. What a word does is not confer the thing. It makes the thing fundable, and arguable, and harder to cut from a budget. That is a real service and a small one, and it should not be mistaken for a discovery.

Nobody was there for the token. Which turns out to be the most important thing about the token.

A Sunday morning at Turk and Hyde, in a mini park roughly the size of a generous living room. Bags, grabbers, gloves. Refuse Refuse has been running cleanups across San Francisco for years and needed nobody’s permission and no software to start doing it.

The work is not photogenic and some of what you are picking up is not litter. You work around human waste in a neighbourhood where twenty-five thousand people live in half a square mile, and the reason is not complicated and does not require a theory: nobody should have to live in that. Not the families in the residential hotels. Not the people in the doorways, who are living in it most of all. Not the kid walking to school through it.

It does something to a person. It is closer to prayer than to the word volunteering suggests, and more humbling. Measured honestly against the other available uses of a Sunday morning in this city — and this city offers many, most of them air-conditioned — it moves you considerably more than pickleball. It is a way of acting in line with a version of San Francisco that a lot of people here can still describe in detail and have not agreed to stop believing in.

You are not there to fix it. You are there because being there is what the belief looks like when it is done with your hands.

Now the uncomfortable word, which belongs to the city and not to anyone holding a bag.

We Live Here opens with a sweep. Police clear a park, Rockey runs, and you are left standing in her tent. A sweep is not merely an eviction; it is a rite with a very old name. The pharmakos was the figure a Greek city drove out in order to be clean — the cleanliness understood as the expulsion.

Set the two gestures beside each other, since both of them are called cleaning.

A sweep removes the people and leaves the conditions.
A cleanup removes the conditions and leaves the people.

One treats the person as the debris. The other treats the debris as the debris, and treats the person as a neighbour who has to keep living there after everyone else goes home. That distinction is not rhetorical and it is not a matter of intent. It is visible at noon, in what is still standing in the park.

At the end of the morning everyone got fifteen SFLuv — a local token, on an app, good at a growing list of Tenderloin merchants. The project describes itself in the flat language of economics: minimise capital leakage, raise the velocity of money within the community. Its volunteer days double as merchant onboarding days, so the list of places the thing works keeps getting longer by one shop at a time.

Fifteen is not compensation and nobody came for it. What the fifteen actually did was route a stranger three blocks to an organic grocery he had never once walked into. A free smoothie. A juice he paid for himself. He will go back, and he would almost certainly never have gone at all.

Which is the same scarce thing the room is after, arrived at independently, by different people, about money instead of attention, on the same four blocks in the same month.

Not immersion. Not impact. A reason to come back. Same park, same hour, next Sunday. One screening is spectatorship and one cleanup is a photograph; the second one is where either of them starts becoming a place. Everything durable in this document turns out to be made of return trips.

And note what SFLuv actually did, because it is the argument of this whole piece happening one layer down: it did not invent a gathering. It found people who were already out there every week for their own reasons, and gave them a rhythm, a list of merchants, and a small circuit for care to run around in. It landed into something.

Immersive work does not have a distribution problem. It has a landing-place problem, and landing places are not built by distributors.

They are built by people who were already doing something else together.

This is the part that will read as unglamorous to anyone whose job title contains the word strategy, and it is the whole finding. A festival screening gathers a room in which the only thing everyone shares is an interest in the medium — which is precisely the room least able to metabolise anything. A poetry night, a mutual aid shift, a library teen board, a Sunday cleanup, a donation-run bookshop: these are rooms with a non-aesthetic reason to exist, and that is the scarce resource.

Service is the engine because nobody has to justify showing up to pick up trash. The gathering is pre-legitimated. The work then lands into an existing reason rather than having to be its own reason, which is the failure mode of essentially every exhibition strategy attempted so far.

And the distribution chain, when it works, has legs. Literally. Cleanup, token, grocery, a conversation on the way to the train, another conversation on Market Street with a man who runs street teams and who — without ever having put a headset on — immediately understood the point, and made a better one back: not every story needs to be told to everyone. Some are for one group and not another. That belongs to the person whose story it is.

Which is, incidentally, the most rigorous statement of immersive ethics produced in this entire document, and it was made on a pavement by somebody who has never been to Venice.

All of which answers the brutal question that gets put to anyone doing this work, usually by someone with a spreadsheet open: why does any of this need to exist when a headset costs five hundred dollars and anyone can buy one?

Because the headset was never the scarce resource. The room is. More precisely: the carefully composed social context is, and there is no version of that you can order for delivery.

Turn that over and the whole economics inverts. Cheaper hardware does not threaten this practice. Streaming libraries do not threaten it. Declining consumer enthusiasm does not threaten it. Every one of those developments removes a distraction and makes the actual function easier to see — which is not access to content, since the content is largely free and has been for years, but the assembly of an occasion. A cinema without a building. A book club for embodied media. A travelling practice for composing encounters, which is a job that has existed for as long as there have been stories and has almost never had a name.

And here is the part that ought to reassure the people with spreadsheets, because it is a genuine win for both sides of this argument.

Consider a man who won an XR hackathon, bought a Quest 3 with the prize money in February, used it once, and left it on a shelf until August. Six months, five hundred dollars, one session. By every metric the industry cares about, a dead unit.

Then two evenings in a room. Twenty minutes inside each time. Hours of talking afterwards.

Nobody sold him anything. Nothing was optimised. The session was the shortest part of the evening and the device spent most of the night in a case on a folding table. And that is the whole business case, if a business case is what is wanted: the practice that treats the headset as the least important object in the room may be the only thing currently capable of making anyone want to pick one up.

The brutal question is: why does any of this need to exist, when anyone can buy a headset for five hundred dollars and download a thousand things onto it tonight?

Here is a Meta Quest 3, bought in February 2026 with the winnings from an XR hackathon. It works perfectly. Nothing is wrong with it. Its owner has used it once since, and it is now August.

That is not a story about a bad product. The hardware did everything it advertised. It is a story about what was actually missing, and it was never the device.

The same person then spent two evenings in a room where the headsets were on for twenty minutes and the conversation ran for hours. And is now, for the first time in six months, thinking about picking the thing up again — but not alone. That is the whole finding, delivered by one dusty box on a shelf.

The headset was never the scarce resource. The room is. More precisely: the carefully composed social context is, and nobody has been counting it as a resource because nobody has been counting it as work.

Which flips the risk model entirely. If the value were in the device, then cheaper hardware, better storefronts, streaming libraries and a collapse in consumer enthusiasm would all be threats. Under this reading they are not threats. They are clarifications. Every drop in the price of a headset makes it more obvious that the bottleneck was somewhere else the whole time.

And what the practice becomes, once you stop describing it as outreach, is something with much older relatives: a cinema without a building. A book club for embodied media. A travelling craft of composing encounters, in which the works are borrowed and the room is the thing being made.

It is also, and this is not a consolation prize, the best available news for anyone who manufactures these devices. A headset used once and shelved is a dead unit no matter how good the panel is. The thing that gets somebody to pick it up again in August is not resolution. It is another person.

Underneath all of it there is a question you can ask in a doorway, with your coat half on.

Did taking it off make the room more legible and human than it was before you put it on?

Notice what that metric does to the business model. Every consumer platform on earth is optimising for session length, retention, return visits, time inside. This one succeeds at the exact moment the device becomes unnecessary — when nobody wants another twenty minutes because the fifteen people in front of them have become more interesting than the rendering.

A technology whose success condition includes its own temporary obsolescence is not going to be built by anyone who needs it to scale. Which is, plausibly, the argument for public money, and the reason this note has an address in Paris as well as one in the Tenderloin.

The medium spent a decade chasing presence. Its mature form may turn out to be about absence: it covers the world, shows you another, and then disappears — and what is left is a trace, and a person sitting across from you whose reality exceeds anything the device was able to say about them, and who, unlike a rendered subject, can talk back and correct you.

That last clause is the whole ethics. Not immediacy. Not authenticity. Contestability. The problem with a representation is not that it mediates — everything mediates — but that it mediates unilaterally, producing an account of a person against which that person has no move.

The conversation afterwards is not the enrichment programme. It is the part where the right of reply gets restored.