I don't think supply is the problem.
It's simply that demand is skyrocketing, but we didn't make software substantially easier in the last decades.
It's simply that demand is skyrocketing, but we didn't make software substantially easier in the last decades.
Most software is complete shit, and we're lucky when the end user doesn't notice too much. As a software engineer, I'm appalled by the average "web app" both in terms of how the product functions and the code making it happen. Accessibility is still a complete afterthought and the documentation on accessibility standards is too confusing.
The only reason many of us are still around is that no one has truly figured out a good replacement for the average software developer yet. The closest we have is things like Wix and Squarespace. The junior and mid level developers of today are ripe for obsolescence and it's a matter of time before someone finally figures out how to give small companies the power of software development without all of the overhead they currently endure.
The world also doesn't need that much software, hence we don't necessarily need more software developers all the time. Software scales with problems to be solved, not with population. Those paying our bills need to figure out just how much of their money they are wasting on what is essentially the same kind of software development that took place 30 years ago.
I think, we didn't scratch the surface of the true demand yet. But, yes, we should get dramatically better in so many ways.
It's only been a few years since developers (outside of these niches) stopped laughing when hearing about sound static type checking, static analysis, model checking, object capabilities, abstract interpretation, message passing APIs, getting rid of NULL, reactive programming, stream programming, let it fail, transactional memory, linear types, etc.
Of course, at the same time, we moved everything to web programming, where everything is broken for different reasons. Research has been studying that for a couple of decades, too, so I figure we should start using more robust paradigms within 10-20 years.
Larger teams also ship faster, there's a lot of complexity to enable that collaboration.
Like what? Earlier systems worked if your computer worked. Adding a server is a DRM move and doesn't help the user with reliability - usually the opposite.
My experience is that in most situations larger teams ship slower, precisely because of the collaboration costs you mention. There are ways to reduce the dependencies, but most organizations can't get there.