AI and programming classes in Banbridge

OpenStreetMap supplies the data: over a rectangle covering Banbridge there are 201.2 km of walkable streets and paths and 30 mapped bus stops. Dijkstra's algorithm gives the walking distance between every pair of stops. Starting from the stop nearest the point OpenStreetMap uses to label the town, Python solves rounds of the nearest 8, 10, 12, 14 and 16 stops three ways: Held-Karp for the exact shortest round, a nearest-neighbour rule that always walks to the closest unvisited stop, and 2,000 random orders for comparison.

For sixteen stops Held-Karp stores 245,761 partial answers and makes about 1.7 million comparisons, finishing in 2.5 seconds, where checking every order would mean about 654 billion rounds. The greedy nearest-neighbour rule is instant and here only 3.3% longer than the true optimum; a random order is more than twice as long. The catch is that Held-Karp still grows exponentially. All 30 Banbridge stops would need about 16.1 billion subproblems, far beyond a laptop's memory. Past a few dozen stops, even exact cleverness gives way to heuristics, and the skill becomes judging how good the heuristic answer is.

A Banbridge teenager who knows when an answer is proven and when it is merely good will not be bluffed by an AI plan, and coding is where that instinct forms. The longer argument is in why teenagers should still learn to code in 2026.