Coding classes in Devon

Each simulated fish knows only its neighbours within a short distance. Every step it applies Reynolds' three rules, in his words: separation, "steer to avoid crowding local flockmates"; alignment, "steer towards the average heading of local flockmates"; and cohesion, "steer to move toward the average position of local flockmates". Then it swims on at a constant speed. The learner starts 150 fish in random places, facing random directions, in a 100 by 100 tank whose edges wrap round, and runs 600 steps. Watching the screen, a shoal forms. But how do you prove it?

The answer is an order parameter, a single number that measures how organised the whole group is. Turn every fish's heading into an arrow of length one and average the arrows. If the fish point every which way, the arrows cancel and the average is short; if they all point the same way, it has length one. For 150 random headings the expected length is about one over the square root of 150, which is 0.082. That gives the learner a baseline, and a way to test each rule by turning it up or down.

Two lessons come out of the table. The first is the sudden switch: between a quarter and a half strength of one small rule, the group goes from half-organised to almost perfectly aligned, a tipping point that no single fish decides. The second is what each rule is for. Without alignment there is clustering but no shared direction. Without separation the order stays high, but 90 per cent of fish end up within one unit of another, crammed together. Only all three together give a shoal that looks and behaves like the real thing.

So a Devon teenager should learn to code in 2026 not to watch simulations, but to measure them. The longer argument is in why teenagers should still learn to code in 2026.