1) Crunching the response numbers to obtain the required EQ to flatten it. Not an easy task within itself. 2) Changing those numbers into EQ coefficients.

I agree that would be tricky. I think I even commented on it, in a previous discussion on this topic. Basically, I showed a screen output from SpectraLAB that looked like a stock market graph. How the heck could a simple algorithm decide the best way to flatten that with only ten bands of EQ?

You know what I'd be happy to see? Something do-able without having to implement complex AI to create EQ coefficients?

Continuous sweeping up and down sine waves output by the DSP, with the screen showing a continuously-updated third-octave analyzer graph of the monitored aux-in using the ratshack microphone.

I sit in the car wearing earplugs and fiddle with the EQ settings as it does this. That way, I choose how to flatten it, but the player has all the features which allow me to do this without dragging the RTA gear out to the car.

Since it's possible to edit the player's EQ by just the numbers along the top of the screen, the third-octave graph could continuously overlay the bottom 3/4 of the screen and not really have much of an ill effect on my ability to tweak the EQ.