Coding and AI classes in Wembley

A rectangle of OpenStreetMap data around Wembley gives the raw material: 223.0 km of walkable streets and paths, 8 stations, 25 parks, 4 libraries, 279 shops and 3,167 mapped buildings. For every building Python works out the walk along real paths to the nearest station, the nearest park edge, the nearest library and the nearest shop. The typical building is 599 m from a station, 506 m from a park and 946 m from a library.

One building dominates another if it is at least as close on every goal and strictly closer on at least one. Any building dominated by some other building is dropped. The survivors form the Pareto front: none of them can be improved on one goal without losing on another.

With two goals only seven buildings survive. At one end is a building 21 m from a station and 193 m from a park; at the other, one 903 m from a station and 8 m from a park. The usual shortcut is a weighted score, say 70% station and 30% park, then pick the lowest. Across 101 different weightings that shortcut only ever picks three distinct buildings, and four of the seven front members are never chosen by any weighting, because they sit in a dip in the front that a straight-line score cannot reach. Add goals and the front grows fast: 44 buildings with a library, 90 with a shop, because with more goals it gets easier to be unbeaten on at least one.

Seeing the whole set of reasonable options keeps Wembley teenagers in charge of choices that AI helps to make, a strong reason to learn to code in 2026. The longer argument is in why teenagers should still learn to code in 2026.