Coding and AI classes in Wellingborough

With one query to OpenStreetMap, the learner downloads every way tagged as a public road class inside a box around Wellingborough: 1,769 mapped ways, from main roads down to residential streets, leaving out service roads, tracks and footpaths. The program turns them into a network whose points are junctions and street ends. The largest connected piece has 2,236 points and 2,572 links. Counting links at each point gives the first picture.

Now the peeling. Round one removes the 811 dead ends. That leaves some junctions with only one link, so round two removes 239 more, round three 90, and so on through eight rounds until every remaining point has at least two links. In all 1,223 points are peeled away. What survives is the 2-core: 1,013 points and 1,349 links, 45.3% of the network. It is the part of the town made of loops, where there is always a second way round. Everything peeled off is a branch: a cul-de-sac, or a whole estate that joins the rest at a single point.

The next step gives the surprising result. There are 1,355 three-way junctions, so a learner expects a healthy 3-core, a region where every junction has three links inside the region. Peeling for three links removes 384 points in the first round, 265 in the second, and after seven rounds nothing is left. The 3-core is empty, and the highest core number anywhere in the network is 2. A junction having three links is a local fact; belonging to a 3-core is a fact about the whole neighbourhood, and the streets do not supply it. One limit to note: roads that leave the box are cut off and counted as dead ends, so points close to the edge are peeled earlier than they would be on a wider map.

Working it through in code is what makes the idea stick, and it is the kind of understanding that keeps Wellingborough teenagers useful alongside AI in 2026. The longer argument is in why teenagers should still learn to code in 2026.