This reminds me of validation loss in machine learning: you push the practice until it gets a tiny bit worse than before -- then you know you had hit the sweet

This reminds me of validation loss in machine learning: you push the practice until it gets a tiny bit worse than before -- then you know you had hit the sweet spot. Applied to everyday things, there is some Zen in knowing that you need this dent to perfection in order to know you had it.