In particular the idea of approximation seems inherently useful. All approximations take the form something = approximation + error, so the idea of having a useful language for the error seems useful. Once you have that language, the tangent line is just a particularly useful example. Only after working with tangent lines for a while would you get to the definition of the derivative.
The nicest thing about the approach is that all of the derivative rules are easy to remember for tangent lines. Product rule? Multiply the tangent line approximations, throw away the high order term. The chain rule? Plug one function into the other. And so on.
The only thing that I found did not flow naturally was L'Hopital's Rule. However that rule is a source of confusion for students who don't understand the importance of the precondition that the limits be of the form 0/0 or infinity/infinity. And furthermore all standard cases where it is useful are obvious upon inspection for anyone who knows Big-O/little-o notation. (Don't believe me? Crack open a textbook and look at the problems that you're supposed to use the rule for. They are all obvious.)
I lack the resources to make or find such a graph myself, so please let me know if something like this exists, or if you intend to create one. My parents are both teachers, and I'm sure they'd benefit from it.
I don't see limits as artificial as all. For example, Zeno's Paradox[1] is resolved thanks to the limit of a succession. And the concept of the infinitesimal, which was something I had thought about before I learnt about limits, is another rather intuitive thing that is explained by limits.
Derivatives are only one of the most common uses of limits, they are definetely not the only reason they are useful.
[1]http://en.wikipedia.org/wiki/Zeno%27s_paradoxes#Achilles_and...
The more you learn, the more you come to realize that the nature of the real numbers is deeply mysterious and not to be treated lightly. That's not to say you cannot convey many useful intuitions to students without the somewhat abstruse formal machinery.
That's more or less what I was trying to say. There are some deeply intuitive concepts for which limits are, at the very least, a very good aproximation. It may occur in some cases that the actual case and the intuition that one has about it doesn't match, but that doesn't mean that the model is arbitrary or artificial.
People understood the solution to Zeno's paradox centuries before we had a rigorous definition of limit. And limiting processes were understood both before, and after, without necessarily referring to limits. (In fact in computer science people prefer the more flexible Big-O/little-o notation for that purpose. Knuth has in fact suggested replacing limits in introductory Calculus with that notation and I'd be quite interested to see how well that would work.)
The idea of infinitesmals were thought up well before limits. In fact the idea of a limit came out of Cauchy's work. Cauchy tried to define an infinitesmal as a function that went to 0. (His idea of an infinitesmal became what we now call a Cauchy sequence.) When limits came they did not explain infinitesmals - they replaced them. It was not until Abraham Robinson came up with non-standard analysis in the 1950s that anyone came up with a sound explanation of infinitesmals and properly explained the connection with limits.
And finally I was trying to explain what things are like from the perspective of the student.)
I think the brain must do some sort of cross training - learn python and improve your C++, so that when I came back to limits they seemed intuitive and not as difficult as I remembered.
1. We introduce limits. Which are blindingly artificial, and so it is hard for us to form a concept about why people want this.
2. We introduce the definition of the derivative. Which instantly results in a practical reason for limits. But typically we make the mistake of introducing the derivative as a function, which means considering the limit of a lot of things at a lot of points.
There really needs to be an intermediate step, which is to introduce the definition of the tangent line. And how to calculate it. Then work up how to calculate tangent lines of various combinations of functions at a single point. And after the person is well and truly comfortable with that, then introduce the derivative.