I learned about the various surface commands and which to use when by experimenting. All that is needed is some time and a computer with Rhino. Sit down, start creating curves and then see what each surface tool does with those curves as input. Change the curves, create new surfaces and see how the surfaces change.
A free resource for introduction to the math used by Rhino is: https://developer.rhino3d.com/guides/general/essential-mathematics/
Ha! Iāve tried at least three different versions of FormZ, and it had many bugs, a few of them pretty serious, including one related to blocks that made working with blocks impossible and this was version 8, so stop claiming non sense.
Pressing each individual icon and observing what it does in the viewport is exactly what I did when I tried Rhino 2 for the first time at the company where I worked about 18 years ago. In a few hours I built my first 3d NURBS house model. Couple of days later and thanks to following the tutorials included in Rhino (the duck, the flashlight and a few others) I was able to build a Pagani Zonda 3d model using 2d blueprints. Pagani Zonda is one of those cars that are relatively easy to model with NURBS. The designer of Saleen S1 also did some clever use of 4-sided surfacing to design the car in Rhino 1.
Modern cars, however, have too many complex shapes (mostly added for the sake of complexity) that are a big challenge to Rhinoās surfacing tools, because they require advanced point editing and surface matching tools, along with proper analysis of surface edge deviation and curvature.
Similar story, very late '90s I downloaded the rhino 1.0 trial, after using autocad for some years (to do 2.5D CAM work), and failing miserably to draw anything in 3D (I still remember the tutorial model). Within about 6 hours I had drawn some semblance of a cylinder head in Rhino and was dreaming of doing 5-axis head porting. Obviously not any sort of advanced surfacing, but as you say, just pushing buttons got me a result, and gave me confidence, very quickly.
Well, FWIW, you donāt necessarily need to have a good grasp of the underlying complex math.
Math was never my thing. Addition, subtraction, multiplication, division, percentages - my interests never went much beyond the basics.
Yet I do some of the most complex surfacing in design, and have millions of parts going through molds. Go figureā¦ 
Is professional level surfacing hard - YES. Do you NEED to understand the math - NO. Maybe that desire is the pinch-point? You may free your mind of the math, if desired. Different strokes for different folks, perhaps.
I remember my last boss and many of my previous coworkers saying we do āmathematical surfacingā ā¦ This is funny, because I never saw them ever calculate something. Not even by code. Most of them couldnāt even code or understood any math involved. You get only exposed to this, when you need to create your own surface tools, and sadly they never understood the real challenge of the job I did for them. I mean you can learn how to read a surface and how it behaves just by pure trial and error. Its really just experience and practise and no rocket science.
Well, you (the designer that is) are doing āit,ā but some coder with a penchant for math ādidā that heavy math lifting for youā¦
Top designers who are also top coders and mathematiciansā¦ Renaissance people, perhaps! (What would Leonardo be doing today???)
Some jobs require you to know many different things on average, others just being a pure specialist. The reason that many of them did not do coding or trying to understand the math however is rather being unmotivated to learn. You donāt have to be top tier expert, but its quite useful to look around the edge. Math books in particular are horrible, but I was just pointing to the fun fact that people claim doing things mathematicaly , but actually having no real clue about. And although I own books about Nurbs, I admit not to fully understand the math involved. 
I smile when I read that with Rhino you can make automotive projects. Why automakers donāt use Rhino instead of relying on software that costs 20 times more than Rhino?
Letās not say heresies!
The main reason is that programs like Alias and ICEM have more advanced tools for control point manipulation, surface matching, edge approximation, etc. If Rhino had the capabilities of the āAlignā tool of Alias (this is how they call the āMatch surfaceā), this single tool would make it perfectly suitable for Class-A surfacing.
A dreamā¦ 
Well thats not a fair comparison. Dedicated software is probably more difficult to learn and I note that even with all that precision, door gaps and hood gaps are still 4-5mm on US models and 1-2mm on high end Jap models.
Rhino is very accessible, its the surface modeling techniques that present difficulty
Itās because they conflate Inches and Millimeters in the model Unitsā¦ 
// Rolf
There are a few reasons for leaving larger gaps between panels:
Hiding imperfections in the assembly is the exact same reason for purposely leaving larger visible gaps between the two halves of remote controls, game controllers, TVās and basically every electronic device made out of plastic and assembled via screws. Mobile phones donāt have that problem, because their plastic back panel is both small and flat.
Off-road vehicles usually have larger panel gaps by design, because their structure is more prone to twisting and bending during driving on uneven roads. As they more, all body panels move slightly, so the larger gaps protects the paint from damaging.
Military vehicles have way larger panel gaps, because their doors must be able to close even if there is some cloth stuck in-between. This is especially critical during emergency situations.
Another one is the size of the panels. Larges panels extend and contract more due to changes in temperature. American cars at least back a few years back in time used to be bigger than European cars.
But yes, I remember my friends in the US how they were impressed by the smaller spaces betwen panels on European cars in the late seventies when I spent some time overseas. And as a European I was never really sure of in which direction all the chunks of American metal was actually going when driving an American carā¦
// Rolf
heh, my daily driven cars are a '91 E30 and a '64 fairlane ā¦ driving the latter is a state of mind, like using the force ā¦ you should hear ben kenobi whispering in your ear and stop trying to have so much control, because the floaty boat does best when you let it do as it likes. 
Agreed. But it took me a near death experience before I fully understood that principle; Follow the car, donāt try to force it.
// Rolf
This car had a 3-5mm gap between bumper and rear wheel archesā¦ It was ugly to see if you knew about it. And ābumpsā on the bumper would accentuate itā¦
Gosh i miss that carā¦
As it turns outā¦
YOU CANNOT MERGE TRIMMED SURFACESā¦ so now I have to rebuild all my trimmed surfaces from the curves of the trimmed surface edge I guessā¦
I am working on an object that has about 10 surfaces, 2 or 3 that are trimmed in complex ways. The resulting model must have NO NAKED edges for calculations like VOLUME:
The underlying geometry has tangential requirements for some edges but not for others. In the real world, the welder will join allthese edges since they are close enough. Tolerance is set very small because the calculations for tangency require this. Unfortunately this obviates the MERGE command.
Join will succeed, but there are still naked edges. Whats needed is FILL GAPS command that just filles in all the holes (which are at the sub mm level) with some rudimentary ājust get it doneā surfacing and thus removes the naked edges.
Going back through the modelling to get surfaces to be adjacent down to the tolerance, is akin to the last 2% of any project taking as long as the first 98%. Its impractical and not a good use of recources.
As hard as it sounds, garbage in garbage out. Rule Nr.1 in surface modeling is to work precise, clean and as simplistic as possible from beginning on to prevent exactly this. You can however model in 10 times bigger scale and scale back down after you are done, which scales error down as well if you still encounter this very often. But never reduce tolerance locally or fill in little surface patches. This is the worst you can do to a model!
