Threnody for the Victims of the Metaverse
Meta is shutting down Horizon Worlds. Everything we built there was real.
There is a ranch on an asteroid. Crystalline fire-pit at the center, floating particles drifting upward like embers that forgot about gravity. Neon-pink hay bales in a loose circle. A barn, a ranch house, a basketball court where people are shooting hoops and talking across six time zones.
Don Carson is sitting in a plush, floating easy-chair near the fire. Former Disney Imagineer, current senior creative director at Mighty Coconut, and tonight he is giving an informal fireside on environmental storytelling to a group of strangers gathered on a crater-strewn rock in deep space. His voice is spatially located (turn your head and it shifts, walk closer and it gets louder). You took a rocketship to get here, launching from the moonbase below, and your brain accepted the transition so completely that the home office you are actually sitting in has ceased to exist.
Nobody is thinking about the headset. They are on the asteroid, hay bale beneath them, fire crackling, listening to a man talk about the philosophy of place the way only someone who spent decades designing rides for Disney theme parks can.
That world is shutting down on June 15, 2026.
Meta is pulling Horizon Worlds from Quest headsets. Mobile-only going forward. Reality Labs has burned through seventy-three billion dollars since 2021. The platform never drew more than a couple hundred thousand monthly active users.
The people celebrating and the people mourning are both missing the point.
We Were Not Building a Metaverse
In January of 2022, we partnered with Voltage Control to try something nobody had a playbook for: take professional facilitators, hand them VR headsets, and see if they could plan, build, and conduct collaborative activities in a platform that had barely left beta. It was rough and messy and thrilling, in the way that first steps always are when nobody knows where the path goes.
By June, we went public. Collaborative Frontiers was a two-week open event with nearly sixty signups, a self-organizing team structure on a virtual moonbase (people walked themselves to Red or Blue platforms, no breakout rooms, no assignment), and a training program called the Collaboration Enrichment Center that consistently took people from zero building experience to creating functional worlds in under an hour. One participant had purchased their Quest that morning (the first gaming console they had owned since an Atari).
The Red Team built a spaceship puzzle game themed after my novel, The Dauntless Gambit, complete with a run-down space barge named Matilda and a facilitator playing the ship’s captain. The Blue Team built a fifteen-room escape complex. When they ran out of time to finish the second half, they crashed the planet into the facility, lit everything on fire, and turned the destruction into the narrative driver. Honestly, it was a better story than the original plan.
And then came the Fall Carnival. Eight concepts we designed to test how people actually collaborate in VR, plus six user submissions that blew past anything we had imagined. A body polling station where people voted Pepsi versus Coke by physically running to opposite ends of a platform (no instructions, no prompting, they just did it).
A wilderness campfire where strangers grabbed flashlights, held them under their avatar’s chins, and told stories until everyone lost track of time. A genie’s lamp where personal questions appeared in colored firelight and people who had met twenty minutes ago were sharing things they would never have shared on a video call. A sailboat retro on an actual sailboat where someone fell overboard and discovered an underwater world nobody had planned for. And Sculptionary, where people stopped performing and started creating, using primitive shapes and buckets of paint, so absorbed in the act of making that the line between virtual and real quietly dissolved.
NPS score of 71. One participant brought their team back to hold their company Christmas party in one of the worlds.
None of this was about VR. All of it was about what VR enabled.
What Outlives the Headset
Five things we learned across those three events. None of them require a headset.
People interact with people, not environments. Every concept where the core interaction was between humans succeeded. Every concept where the interaction was between humans and props fell flat. Rose Thorn Bud, a proven retrospective framework, completely lost its power when the sticky notes were virtual objects floating in a park. But the campfire, the genie, the body polling? Those worked because the content was each other.
“Play disarms fear.”
Paul Tomlinson said this during every session, and he was right. Sculptionary proved it. The body polling proved it. When people play, they stop performing anxiety and start being present. The facilitator’s real job is not managing the process. It is creating the conditions where play becomes possible.
The environment is the facilitator. The Moonbase self-organized teams without a single instruction. The genie’s lamp triggered engagement without prompts. The body polling station needed zero explanation. The best facilitation is the kind you do not notice. You think you are choosing freely. The room already chose for you. (Cicero codified this as the method of loci two thousand years ago. We just rebuilt it in polygons.)
Suspension of disbelief is a storytelling problem, not a technology problem. The Epic of Gilgamesh transported people to other worlds four thousand years before anyone strapped on a headset. Narrative is the technology. VR was a delivery mechanism. So is AI. So is a campfire and a voice people trust.
And the one that still gets me: memory does not distinguish virtual from real. Ask anyone who attended those events and they will tell you about standing on the asteroid. About the fire. About who was next to them and what Don said about placemaking. They will not mention the headset. They will not mention their living room. The experience was virtual. The memory is as real as any conference they have ever attended in the flesh.
That is the part people miss. Not the polygon count. Not the frame rate. The memory.
Seventy-Three Billion Dollars and a Couple Hundred Thousand Users
VR failed commercially. That is not a debate. Meta spent seventy-three billion dollars on presence technology and never once figured out that the thing people wanted was not a headset. It was each other.
Now, we read the schadenfreude from people who never built anything in it, the “told you so” crowd treating this as proof that presence and connection were always the wrong bet. I disagree entirely and have the receipts to prove it. This isn’t the first time that something of real human value was cut because the financial ROI wasn’t there:
Google Reader died and people mourned it for years because it was not just software, it was a way of thinking with the internet.
Vine died and took an entire native grammar of comedy with it.
Google Stadia died too, and for a while it had given people a real glimpse of what frictionless play across distance could feel like.
Even the Sega Channel didn’t last. I know. I was a beta tester in mid 1994.
But VR did not fail because presence and connection are not valuable. It failed because a headset was the wrong vehicle. The destination was right. The horse was wrong. This is the recurring category error: a vehicle fails, and people declare the destination imaginary. They do it every time. A bad headset does not invalidate presence any more than a failed train line invalidates travel.
That mission is ancient. The Pony Express carried it. The telegraph carried it. The telephone, the workshop, the video call, and now AI. The medium keeps changing. The mission never did. One horse hands the satchel to the next. One wire hands it to a switchboard, the switchboard to a handset, the handset to a glowing rectangle, and now the rectangle to a model. The relay keeps changing shape, but the message is still the same human ache to reach across distance and do something together.
AI is doing now what VR tried to do: collapse the distance between intent and execution, between people and their ability to create together. VR attempted it through simulated presence. AI is doing it through collapsed translation. Different medium. Same letter. And the letter still needs to reach Josephine. She never cared about the horse.
I have made this mistake before. I have fallen in love with the container and mistaken it for the cargo, spent precious energy mastering the stage instead of protecting the thing worth carrying across it. You pay twice for that error: once when the medium resists you, and again when it disappears. I knew that VR was likely to fail during this iteration, but if I didn’t try, then who would? And if not then, then when?
The Worlds Will Be Gone on June 15
I built worlds on Horizon Worlds. I themed a spaceship after my own novel. I watched strangers become friends around a crystalline fire on an asteroid while a former Disney Imagineer talked about the philosophy of place. I watched someone who had never owned a gaming device more recent than an Atari learn to build in VR in under an hour and make something they were proud of.
Those worlds are being deleted. The asteroid ranch, the carnival grounds, the Collaboration Enrichment Center, the Matilda. The moonbase where people organized themselves by walking to colored platforms. The campfire where strangers held flashlights under their chins. All of it.
There is a particular heartbreak in building something beautiful on rented ground. You can do everything right, make something people genuinely love, and still wake up one morning to find out a product manager in Menlo Park has functionally set fire to your town. That is the lesson: build where you can, love what is real inside it, but never confuse borrowed infrastructure for permanence.
But the connections were real. The learning was real. The joy was real. And that is not sentiment (it is the science of how memory actually works). The experience was virtual. The memory is not.
The fire on the asteroid is going out. The hay bales will disappear. Don’s voice, drifting across a ranch on a server somewhere in California, will go silent.
Seventy-three billion dollars could not make the medium last. Under an hour was enough to make the memory last. The person whose first console since an Atari was a Quest will not remember Meta’s burn rate. They will remember the asteroid.
You built something real on something temporary. So did I. So does everyone who makes anything on a platform they do not own. The question is not whether the medium survives; it never does.
The question is whether the thing it carried was worth carrying; it was.
Bye for now.