Vibe coding and AI agents classes in Kilmarnock

The learner downloads every building mapped on OpenStreetMap across Kilmarnock, 12,925 of them, and treats each as a job for the team: check this building. Each building's ID is turned into a large number by a hash function. The modulo method gives the job to agent number "hash modulo team size". The ring method places each agent at one or more points on a circle of hash values; each job goes clockwise to the next agent point it meets. Adding an agent only takes over the jobs just behind its new points.

With modulo, adding a fifth agent moves 80.6% of the jobs, even though the new agent needs only a fifth of them. Going from 8 to 9 agents is worse: 88.5% move. The ring moves close to the ideal share, but with only one point per agent the circle is carved unevenly: with four agents the busiest had 1.61 times an average load and the quietest 0.29 times, and with eight the quietest was down to 0.08. Giving each agent 100 virtual points smooths the circle: 19.2% of jobs move when growing to five, 11.5% when growing to nine against an ideal 11.1%, and no agent carries more than 1.15 times the average.

Designing for change keeps Kilmarnock teenagers in control of the agent teams they build, and learning to code is where that design sense comes from in 2026. The longer argument is in why teenagers should still learn to code in 2026.