This is a pretty good introduction. I highly recommend Feynman's book "QED: The Strange Theory of Light and Matter" as an excellent in-depth work that does not sacrifice accuracy for the sake of making difficult ideas understandable. It is both very clear to the layperson and accurate to the physics.
Wholeheartedly seconded. This book occupies a nearly unique position in the physics literature: it is neither a textbook nor a popularization. It assumes little more knowledge (of math or physics) than the typical popularization, but it explains what is very nearly the true, complete structure of its subject matter (quantum electrodymanics). Now, the methods that it teaches are absolutely unwieldy: it would be hopeless to do any real, meaningful calculation by drawing countless little arrows! But (as I think Feynman says) you can go to grad school to learn the efficient tricks. The underlying concepts will carry through essentially unchanged.
An interesting and surprising fact about Feynman's books is that he never _wrote_ any of them, as such; he famously disliked writing. QED is essentially a transcript of a lecture (although I don't know how much polishing and editing was done by Feynman himself; probably some). Same with his collection of physics lectures. His most popular book, "Surely You're Joking, Mr. Feynman", was transcribed from a series of interviews, pretty much verbatim.
Relativity: The Special and General Theory by Einstein would also fit that description, I think. (Though I haven't read Feynman's book so I may be misunderstanding you.)
It's a marvelous book, but a little frustrating as well for not quite explaining enough to calculate with, even super-inefficiently. Can't we get a correct formula for the propagator, including the four spin components? When I tried to fill that in from other sources I was snowed under by all the formalism or prerequisites -- yes, it's probably a short step if you already understand it or you're smarter than me.
If someone made a clear-but-impractical QED simulator using, say, lattice gauge theory, I pledge to sing its praises. (You'd expect this to be possible for general relativity too.)
It would be nice to have a version that uses complex numbers and linear algebra instead of spinning arrows. A little bit more math would make it easier to connect with the "grown up" version of the theory.
It starts off with a spring mattress analogy (like the one in the linked article, but with more math) and goes on to more advanced material from there. I remember it requiring little background besides LinAl and multivariable calc.
It's been a while but I'm pretty sure the spinning arrows are Phasors. Ironically the spinning arrows might make more sense than the math that describes them.
Much less famous than Feynman's Messenger Lectures at Cornell (or his real "Lectures" that got turned into the books) are the lectures he gave in New Zealand that became QED: The Strange Theory of Light and Matter.
