Back to square one, again

While I'm perfectly able to compile some yocto-based image, and to
run it in qemu, I'm basically unable to go beyond that.

This stressful situation is due to a few factors.

From one side, my task is not clear at all.  I've received some
contextual information, and I've got a very high level idea of what
I should accomplish.  The task definition itself is full of question
marks and "this is yet to decide" sentences.  Everyone is very kind,
but also quite confused.

I'm supposed to refer to a colleague, K, who is working in a different
team, by the way.  He is the one who decided for Yocto (the initial
choice was Buildroot, I've heard.  Buildroot would be way simpler
to digest!).  He is also the one that has a better knowledge of the
Yocto system.  I've contacted him preliminarily but, beyond that,
it is not easy for me to come up with a sensible question.

I've got a generic direction where I want to go.  Specifically, I'd
like to end up with a /dev/uioX device in my qemu emulation, bound
to some arbitrary memory area.  That's the best approximation of a
real goal that I could come up with.  I'm also missing all the basic
tools to get there.

By now I've figured that the current repository, where my team and
K's team commonly work, is not the right place where to start.
There's already a few things going on there, and this makes it more
difficult, in the same way as you can't have a clear vision while
people around you are working and there's dust everywhere.

Let's go back to square one, check out a 'vanilla' yocto, build it,
and do my thing without the choir.  Hopefully, this will unlock
enough knowledge, and allow me to understand what my mates did and
what they expect from me.  If anything I'll be able to ask relevant
questions.

All of this is horribly slow.