Coding and AI classes in Withington

The learner downloads the 2021 census count for every output area in Manchester and the official centre point of each, then keeps the 386 areas whose centre falls inside a 5 km square around Withington. Added up by us, those areas hold 141,227 people. Python drops each area's population into a cell of a quarter of a square kilometre: first squares 500 m across, then hexagons of the same area, whose centres are 537.3 m apart. Then comes the real experiment: the grid is shifted by a random amount and the binning repeated, 200 times for each shape.

Hexagons do what is claimed for them. A hexagon's corner is 310.2 m from its centre against 353.6 m for a square, and all six neighbours sit at one distance, where a square has four close neighbours and four diagonal ones 707.1 m away. But neither shape fixes the placement problem. Slide the grid and the population of the busiest cell swings by about half its typical value for squares and hexagons alike, and the busiest cell itself jumps around by some 870 to 895 m. A map that announces "the densest spot has 5,300 people" could just as honestly say 3,300.

So the shape is a refinement and the placement is the bigger choice. Careful analysts show more than one placement, or smooth the data, or at least say how the cells were laid down.

A Withington teenager who has coded the bins knows a map is an argument, not a photograph, and can hold an AI's chart to account. The longer argument is in why teenagers should still learn to code in 2026.