(group 1) received an error message when they wrote code that failed to execute properly: ‘You have failed. Please try again.’ The other half (group 2) got a slightly different message: ‘You have failed. You’ve lost 5 points. You now have 195 points. Please try again.’ Everything else about the two groups was identical. This small distinction made an astonishing difference. Group 1, on average, made twelve attempts to solve the coding puzzle, and had a success rate of 68 per cent. Group 2, on average, made just five attempts to solve the puzzle, with a success rate of 52 per cent. The first time I heard about this experiment I was astonished. Purely because there was an arbitrary, meaningless ‘penalty’ of five points for failure at the puzzle, the 25,000 people in group 2 (from all around the world) made, on average, less than half the number of attempts at the puzzle than those in group 1. As you might have guessed, Rober’s interest wasn’t really in teaching people to code. He was most interested in how we think about failure.