I couldn't help but notice a resemblance between what you are describing and a lot of machine learning algorithms.

It seems like the most successful ML algorithms are self-correcting ones, rather than ones that attempt to calculate the exact correct answer in one go.

Coincidence?