The MapCannon Lab
MapCannon is playful at the surface, but the decision underneath should be explainable. These notes cover the geometry, filtering, and failure boundaries that turn a surprising point into a useful travel starting point.
A rectangle gives every latitude the same vertical space even though the Earth gets narrower toward the poles. MapCannon derives global latitude from a sphere-correct distribution before filtering to land, so a broad “anywhere” throw does not quietly over-sample polar regions.
These are different probability spaces. A country picker can give each country one chance; a coordinate picker gives area more influence and still needs an explicit land or place filter. MapCannon labels the constraint instead of calling a filtered result universally random.
A landing is useful before every nearby-place query finishes. The app resolves a destination, shows a destination-only card when enrichment is sparse, widens the search when appropriate, and asks you to verify the map before traveling. A provider timeout should not erase the decision the cannon already made.
Selected polygons are clipped to bundled land topology where possible. Rejection sampling keeps a dart inside the chosen boundary, while a known interior anchor prevents empty or pathological shapes from producing an ocean-only result. Deterministic seeds are available for offline experiments, never as a hidden production identity.
Fire the random destination generator or read the data methodology. The Lab describes the selection model; it does not pretend that a filtered land result is the same thing as an unconstrained point on Earth.