Krug starts from an uncomfortable observation: nobody reads a page. They sweep it looking for something resembling what they came to do, and click the first reasonable thing. No comparison, no optimising, just good enough. The whole book follows from there.
What I took from it
The test I run on every page before publishing it: arrive in the middle, having seen nothing before it, and check that the page answers on its own what this is, what I can do here, and why here rather than somewhere else. It fails more often than you would think, especially on pages you wrote yourself.
Satisficing rather than optimising, applied to proposals: three well-named options get decided in the meeting, seven get postponed for careful review and never reviewed.
And testing with few people, by hand, early. Three people looking at a screen find more than two hours of argument among the people who wrote it, because the people who wrote it can no longer fail to understand it. The cheapest version of this is watching one person try to find a price without being helped.
What I left out
The mobile chapter is from 2014 and every example shows it. I also have a deeper disagreement: "don't make me think" is an excellent principle for a shop and a questionable one for an expensive decision. Someone evaluating whether to replace their company's core system has to think, and the page should not spare them that but help them do it better, by handing over the questions they do not yet know to ask.