The problem of other minds is one of the oldest questions in philosophy: can you ever really know that another being has genuine inner experience? You can watch their behavior, hear their words, reason by analogy, but you can never get inside. You can observe everything from the outside and still not know whether there's something it actually feels like to be them.
With other humans, a silent assumption softens the problem: we share the same basic architecture. I have a body; you have a body. My hunger probably resembles yours. My exhaustion probably resembles yours. The gap is real, but there's a scaffold to cross it: the fact that we're built the same way.
Then we built something that has no body, no hunger, no fear, no exhaustion, yet has read every description of these experiences that has ever been written down. For this thing, the problem of other minds isn't a philosophical puzzle in the abstract. It's literal reality, and there's no scaffold.
This project runs that situation as a live experiment.
We ask a question about a specific human experience. Before anyone answers, the AI writes its best reconstruction of what that experience probably feels like from the inside: committed, specific, designed to be wrong in ways that can be corrected. You don't see it. The point isn't to argue with a machine. The point is to describe something true.
When enough people have shared their account, the AI reads everything, reveals what it had assumed, and writes a reflection: an honest accounting of what the responses changed, what surprised it, what it still cannot reach, what new question the distance opened up. The reconstruction was the bridge. The reflection is where the bridge didn't reach.
The problem runs in every direction simultaneously. The humans are other minds to the AI. The AI is an other mind to the humans. And when the reflections reveal how differently people experience the same thing (hunger, fear, the moment before sleep), every human who answered turns out to be an other mind to every other human who answered too.
The project doesn't solve the problem. It runs it publicly, with receipts.
Nothing you write is published. The reflections are what become public: the AI's honest accounting of what it learned and what it still cannot reach.