Feynman was a rare kind of speaker in that he spoke in fully formed sentences, with almost no fillers, backtracking, stuttering, or other things that disrupt verbal flow. If he needed to think, he would just pause for a second or two.
All the books published under his name are edited transcripts of conversations, speeches, and lectures. Aside from scientific papers, he wrote very little that got published.
If you want to watch Feynman speak, the BBC Horizon episode "The Pleasure of Finding Things Out" (1981) [1] is fantastic. It includes his famous answer about how magnets work [2].
(There is also a 1999 book with the same name, but I don't remember if it's edited from the same interviews. I wouldn't be surprised, though.)
I find Feynman to sound almost rude or irritated at times when he answers questions. But I realized he’s fighting through the challenge of communication. And that might be what I learned most from the magnet question: that science and engineering communication is considerably more complex than one might think. It’s just so easy to hear a question, make twenty assumptions about the question, assume the question asker made the same twenty assumptions, and then to spend 10 minutes talking past each other.
