I'm constrained by my ability to learn and that is pretty much it.
Well, I think it is more subtle than that. You are relying on certain externalities - for example, you can't build a chip fab without spending an enormous amount of money, and you need all that power so you can work in Python and not ASM, you can't lay ubiquitous bandwidth without spending an enormous amount of money, etc.
What you are doing is equivalent to your friend creating whatever she can imagine in MicroStation (or whatever CAD architects use these days). It's just that she seems to be more aware of the commercial aspects of actually making it happen than you are.
You're right, but these externalities are accessibly cheap.
Last month, my website had ~150,000 visitors. People who I know and look up to in the tech community tweeted about it (it was for an article I wrote).
I was able to design something, implement it (the software the runs the site [on top of the OSS stack]), then release it to the world and watch people use it.
It wasn't just a demo, it was a real-life, functioning product.
I pay $20/mo for hosting on linode. $20/mo gets me the ability to compete with multi-billion dollar media publishers, and multi-billion-dollar software companies (faecbook, twitter, etc [obviously my $20/mo doesn't scale. The point is that I'm creating the same thing they are, just smaller).
I got "noticed" by a whole bunch of people who used my product.
This is not the case with architecture. You could make an really cool model, or design in CAD, and nobody outside of your social circle would likely see it.
I just paid $48 for a year of hosting on prgmr because I wanted to give some friends of mine a sandbox to play in while I teach them python.
Imagine an architecture teacher giving their student a plot of land, as well as a crew of builders, and a yard full of raw materials.
That's where OSS software is right now, and that is amazing.
Well, I think it is more subtle than that. You are relying on certain externalities - for example, you can't build a chip fab without spending an enormous amount of money, and you need all that power so you can work in Python and not ASM, you can't lay ubiquitous bandwidth without spending an enormous amount of money, etc.
What you are doing is equivalent to your friend creating whatever she can imagine in MicroStation (or whatever CAD architects use these days). It's just that she seems to be more aware of the commercial aspects of actually making it happen than you are.