Girard's diagnosis
Desire is learned from models. Rivals become each other's doubles. Crowds end the crisis by choosing a victim. We take the diagnosis and refuse the conclusion some of his readers drew, that order needs an enemy.
An arts lab for dialogue across difference
antimemes.is builds instruments for seeing the culture you're inside, meeting the people on the far side of it, and getting credit for understanding them. Not for winning.
The map behind this text is live. Each face is a public figure, linked where their published words on AI agree. Drag one. Tap a face to read them.
We want what our models want. Online everyone is a model and everyone is a rival, so wants and grudges spread the same way: by contact.
>be me
>no opinion on open-weights AI
>watch someone I admire dunk on it
>now I have a strong opinion on open-weights AI
>it isn't even mine
On the map, the links are real agreements. The spread is a simulation.
Gregory Bateson called it schismogenesis: two groups define themselves against each other, every move answered by its mirror, until the difference is all either side has left.
Feeds run it at scale. Sort everyone by one question and the middle disappears.
Loading the question…
When rivalry has worn away every other difference, crowds settle it the oldest way: all against one. René Girard called it the scapegoat mechanism. Online it never finishes. There is a new one every morning.
The node in the middle is nobody. That's the point.
Every move below is a reasonable thing to do. Each one hands the rivalry a new object.
Now there's a winner and a loser, and a highlight reel of the loss.
Now there's a referee, and both sides can fight the referee.
More models to imitate, faster. The feed doesn't care which side you're on.
Congratulations, you found today's scapegoat.
Give people something to want that the other side has to grant, and that never runs out.
A take spreads in a second. Having listened to someone doesn't spread at all. Understanding has to be done again, by each of us, every time. That's the opening this lab works in. Call it anti-mimesis: meeting difference without imitating it or fighting it.
Same 107 people, sorted by what they actually hold. The dashed lines are descriptions each camp could write of another, waiting for the described side to sign. Ringed faces sit between camps: the translators.
Desire is learned from models. Rivals become each other's doubles. Crowds end the crisis by choosing a victim. We take the diagnosis and refuse the conclusion some of his readers drew, that order needs an enemy.
People can coordinate by reaching understanding, not by force or price, when what they say has to answer for itself. The lifeworld, the shared background we all talk from, is what feeds keep colonising. We build for the lifeworld.
Not agreement. Meta-consensus: every group able to describe every other group in words the described side would sign.
Draft description, written by our pipeline from public words
Nobody in this camp has reviewed it yet. Only they can sign it. That is the work.
The lab makes tools, and every tool has to produce something you can point at.
Live: The Great AI Debate
Public figures are placed by what they've said on the record. Every claim is tied to the exact words and a link to the source, checked in code. Groups are found in the data first and named afterwards.
Next
In an anti-debate nobody rebuts until they can state the other side's case so well the other side agrees with it.
A clean, mutually understood disagreement counts as a result.
Coming
Like Community Notes, pointed across groups. A note about a group earns two things: fidelity, which only that group can grant, and portability, when others find it clear.
First release, and the first instrument: See
Where do you stand on AI's future? Answer 27 statements with the names hidden. Then see which of 107 public figures you stand with and against, quote by quote, and find your place among the camps in the 3D Lifeworld Atlas.
debate.antimemes.is is part of this site: same map, same rules, one level down.
Pick a face to read what they said.
Working labels made by clustering. The people named should replace them.
How the map feeds the rest
SeeLive now
The camps come from public figures' own words. Your answers drop you among them. Say you land in the camp called one camp, the ringed marker.
MeetNext
From there, read the strongest case of the camp farthest from you on the map, another camp. Then write a description of them that they would sign.
MoveComing
Only the described side can sign. A signed description earns bridging reputation, Community Notes style. Agents can draft on a separate board and never sign.
Meta-consensus
Not agreement. Each dashed line is a description waiting for the other side to sign. Signed so far: 0 of 12.
People are placed only from what they've published, and every placement can be disputed.
We infer nothing about private people. Your answers stay in your browser unless you choose to share them. Privacy
We quote a sentence or two and link to the source. Full transcripts are never republished.
Software can draft and assist. Only people in the described group can sign a description of it.
We don't score who's right. We score whether groups can describe each other.
The survey is for people 18 and over.
Where this stands today: each figure is read by machine from about three public works, and none of them has reviewed it yet.
Take the survey and find your place on the map. Then read the strongest case of the camp farthest from you. The hard part comes next: describing them well enough that they'd sign it.
Public figure
Read by machine from public works. Not reviewed by them. Full profile on The Great AI Debate. How to report a misreading.