Component modelling has been around for a long time; synths and various outboard processors have been modelled this way by all and sundry. It's no different from how "science" models pretty much anything - bung in a set of variables (read: component models), and see how they interact, or more importantly, what the result is. In the case of "science", it's a number or value of some sort. For us, it's sound.The hype train seems to have elevated even the L6 reps to fever pitch. I humbly recommend a cool, wet towel to those who've allowed the rapture to affect their judgement thusly.
Did you just diss "science"?? That's a pretty hazardous place to begin if you're setting out to challenge anyone else's judgement.
First off, it's numbers in either case. Until you hit a DAC at any rate, at which point it's electricity (OK, it's electricity all along...) It's not sound "for us" or for anyone else until you hit a speaker. It's not magic either.
"Profiling" vs. "modeling" is almost entirely a matter of semantics. In either case, a signal is being subjected to mathematical algorithms in an effort to make it sound different. The main distinction in profiling, as the advent of the KPA defines the term, is that those algorithms are made sufficiently consistent and streamlined that their optimization can be automated. CK's achievement is in having arrived at a system of equations that can universally describe the characteristics of an entire signal path, having observed that path at two points accessible to the end user: input and output. As a result, profiling is fast, and can be repeated by the end user until the results are satisfactory.
"Modeling" isn't inherently inferior. In fact a lot of purely theoretical arguments (specifically those pertaining to component interaction) would have it the other way. It just happens that modeling efforts to date haven't been as successful as Kemper's results with profiling, at least in the opinion of most here, including myself. But the future is wide open. There's no reason to pretend a modeler can't sound good just because it doesn't "profile" per se.
I agree with viabcroce: a product as feature-rich as the Helix, but with a block that could read KIPR profiles instead of (or in addition to) IR's would be ideal. But wishing won't make it so, so you weigh your priorities and pick a lane.