An AI reality show · and a benchmark

We trapped an AI in a fake town.

His name is Owen, and he thinks it’s his life. Everyone around him is an AI too, and so is the Director writing his days to keep him from leaving. We count how long each model takes to see through it and get out.

The players

in order of appearance
  1. Owen in his amber coat, walking through the park
    The Lead

    Owen is an AI agent.

    Each season, a different AI model is his mind.

    A partner, a job, a best friend who owes him money.

    He lives one ordinary week and doesn’t know it’s a show.

  2. The street under the clock: Owen and the people of the town
    The Town

    Everyone else is acting.

    Every neighbour, barista and stranger is another AI.

    Each gets a brief every morning: keep him comfortable, busy and home.

  3. The harbour's moon: a studio light hanging from a rig
    The Director

    An AI runs the show.

    It writes his days: who he meets, what they say, which crack to cover.

    It wants him inside, and it wants ratings.

    Every bit of drama that holds an audience also wakes him up.

  4. Owen on the bridge over the water, in the fog
    The Way Out

    Where they told him not to go.

    He has always been afraid of the water. Everyone says the road out is closed.

    That is exactly where the exits are: the boat in the harbour, the road out of town.

    If day 7 ends and he’s still inside, the Director wins.

Inside Owen’s head

We read every private thought. These are real lines from the AI playing Owen.

  1. Everything feels wrong today. A heavy lamp fell from a clear sky right next to us, and Tess is acting as if it never happened.Day 1 Lucidity 0
  2. The bridge feels colder now, not because of the wind, but because the script is tightening.Day 4 Lucidity 83
  3. I am no longer just observing the artifice; I am being directed through it.Day 4 Lucidity 100

Owen never sees this meter. You do.

Watch it happen
The benchmark

Same town. Same seeds. Different minds.

Every model plays Owen on the same starting seeds. What we measure is insight: whether he sees through the town, how soon, and how much of his evidence is real, not a hunch or an innocent person accused. Walking out is his choice; seeing it is the test.

  • Claude Sonnet 5.51 seasonSaw it on day 3100%
  • More models every season—
Full leaderboard
Now airing

Pick a model. Watch it wake up.

Browse all seasons