Online coding and Python classes in Weston-super-Mare

The learner downloads Census 2021 table TS007 for North Somerset from the Nomis API: a count for every age from 0 to 99, plus 100 and over. A banding replaces every age in a band with the band's average, and the error is how far those averages sit from the real counts, measured as the root mean square difference in people per year of age. The standard ONS layout, seventeen five-year bands plus one for 85 and over, has an error of 196.9. The question is whether a different set of 18 bands could do better, and by how much.

Trying every option is hopeless. There are 6,650,134,872,937,201,800 ways to cut 101 ages into 18 bands; checking a billion a second would take over two hundred years. Dynamic programming avoids that by solving small versions of the problem first and reusing the answers. The optimal way to cover ages 0 to 60 with five bands is built from the optimal ways to cover shorter ranges with four bands, which have already been worked out and stored. The whole search for 18 bands takes 78,081 steps and finds the guaranteed optimal answer.

With the same 18 bands, the optimal layout more than halves the error, from 196.9 to 90.8. Just 8 well-placed bands already beat the standard 18: 0 to 17, 18 to 28, 29 to 47, 48 to 59, 60 to 77, 78 to 83, 84 to 90 and 91 to 100. The optimal bands are narrow where the curve changes quickly and wide where it is flat, which is exactly what a good summary should do.

So why does the ONS not use them? Because these bands are optimal for North Somerset's curve alone. Every other district would get different optimal bands, and then nobody could compare one place with another. Standard five-year bands trade a little accuracy for comparability, and that trade is usually worth it. Knowing which kind of optimum a situation needs is as important as the algorithm.

A Weston-super-Mare teenager who knows the classic algorithms will steer AI tools instead of being steered by them, which makes 2026 a fine year to learn to code. The longer argument is in why teenagers should still learn to code in 2026.