"How a program works and how it performs are very different things." -- Yes, yes... That's also the case for many random x and y.
"By all means, use your brain and your training to
guide you; it’s good to make predictions. But not all predictions are good." I didn't find a translation for "La chèvre et le chou", which means keep the goat and the cabbage and is ironic. So yes, predictions are sometimes good sometimes not that good. A very many random x also comply to this proposition.
And so ooyawnnn.
But to have a little positive note, some quotes are fun, I actually enjoyed this troubling Mise-en-abîme "I once generalized from a single data point, and I’ll never do that again!"
 Not lucky today, it seems to not have any English translation either, interested reader can wikipedia it or tell me wrong.
It's hard to remember to practice those platitudes day-to-day, which makes them worth repeating. You must register your predictions beforehand in order to avoid hindsight bias. You also have to avoid getting attached to them. Programmers very often commit map-v-territory mistakes, asserting how a thing works as if it were an observation of how it performs, and so on.
If all this is old hat to you, well, congrats! But it's not to everyone. And keep reading. It gets more practical after Ch 2 or 3.