Some part of this is that Kay is more concerned with cognition and learning, and especially in children, than he is with computer programming as such. Programming happens to be a uniquely exciting way to engage with mathematics, and computers happen to be wonderful tools in many types of learning, so he has devoted of his career to inventing programming environments and interfaces for kids.
Because all programming – even to some extent pure numerical algorithm design/optimization – is an exercise in communication and human-usable abstraction, most programmers should probably spend more time on developmental psychology than they do currently.
I think he understands many things we'll only recognize much later, and one of these is that kids need clarity. (Not simplicity; people talk down to kids all the time.) And clarity is orthogonality, it's structure, it's everything we need in our funny little industry and nothing we have.
There are 20 people I admire in this field and the first seven are Alan Kay.
Because all programming – even to some extent pure numerical algorithm design/optimization – is an exercise in communication and human-usable abstraction, most programmers should probably spend more time on developmental psychology than they do currently.