Coding classes in South Ayrshire

Every number in the model is an assumption, labelled as one. Seawater is taken to hold 35 grams of salt per kilogram, a typical figure for the open sea. To get one tonne of salt, the pan must take in about 28,571 kilograms of seawater and boil off 27,571 kilograms of water. The water first has to be warmed from 10 degrees to boiling, and then turned into steam, which needs far more energy per kilogram than the warming. The learner adds the two parts, divides by the energy in a kilogram of coal, taken as 25 megajoules, and allows for a pan that turns only a fifth of the coal's heat into useful work.

Then the sensitivity check. The learner moves one assumption at a time between a low and a high value and records how much the coal changes. Pan efficiency, from 10 to 30 per cent, swings the answer by about 19.5 tonnes. The energy in the coal, from 20 to 30 megajoules per kilogram, swings it by about 6 tonnes, and the saltiness of the sea, from 30 to 38 grams per kilogram, by about 3.7 tonnes. The temperature of the incoming sea, from 5 to 15 degrees, barely matters: less than a quarter of a tonne. So if a historian wanted a better estimate, the first thing to find out would be how efficient the pans were, not how warm the sea was.

The last row is the bug. Salinity is often quoted as 3.5 per cent, and typing 3.5 into code that expects grams per kilogram silently makes everything ten times too large. Two simple checks, that the seawater must weigh more than the salt and the water boiled must weigh less than the seawater, will not catch it. A better test compares the answer with a rough figure worked out by hand, which is the habit this project builds.

Knowing which assumption drives an answer is why a young person in South Ayrshire should still learn to code in 2026. The longer argument is in why teenagers should still learn to code in 2026.