Choose a restaurant on the map. You only observe your own dinner; a mysterious policy chooses in parallel. The orange bars are uncertainty: they shrink only where evidence has been spent.
🙂you
🤖 policy
dinners
0 / 40
current leader
—
your regret
0.00
policy regret
0.00
last observation
Start at home. One dinner is evidence with noise, not the restaurant's whole truth.
A wide orange bar means: this place has not been sampled enough to know much yet.
The mysterious policy is Thompson sampling. It samples one plausible quality from every restaurant's current uncertainty, then picks the best sample. Unknown places keep getting chances; known-bad ones fade away.