With a painting you are getting compared to every other painting, with a program you are compared to other programs for the same task.
Maybe I hail from an alternate universe, but on my planet, everything that's obviously worth doing in software is already being done by 5 different startups and 20 open source projects (I'll grant that it's often done badly, though). How do you find "tasks that no existing software does"?
On my planet, I'm learning to settle for "doing better", or "doing more fully", or "being more accessible in environment X".
You can pretty easily find an area where you can easily be the only/best open source tool for the job and there are a lot of people who will use your tool just because it's open source.
And to be honest, there are actually piles and piles of money just waiting to be made this way.
We (as an industry) are trying to invent flying cars and the world's 482,382nd social networking app with zero revenue model and other crap crap like that.
Meanwhile, your dentist's office is using software that sucks, probably costs $20K a year, and looks like Microsoft Office 2K3 threw up all over their computer screen. But we ignore that, in favor of creating yet another new Javascript framework-of-the-week or something else the world doesn't even remotely need.
I'm currently working at a privately-owned company that makes software for one of these unsexy niches and they're making very solid money. They're literally the only one making software in this space. It will never be a billion dollar space, and probably never a $100mil space, but there is a need and a market of companies willing to pay well for this software.
Follow up with that we get bunch of startups and tools from developers for developers because that is what we focus on. Where making small tool for some other profession could be even more interesting than trying to write better left pad or another package manager for javascript.
But also people in different fields have to use some stuff from 10 years ago or god forbid from previous century like 90's and you don't want to get dirty.
but maybe that's natural for such a young industry?
Usually you are compared to other paintings done with the same technique and/or style. No one is going to compare your impressionist watercolors with abstract geometric acrylic painting. So in a way it's the same. There are categories both in painting and in software development.
No, it's damn hard.
In case you need examples:
I can't find a good surface subdivision modeler for Android tablets.
The desktop clients for Upwork are a bloated mess.
The Field Nation mobile client crashes twice a day for me, and both the mobile client and desktop website present a massive amount of information in a cluttered, completely disorganized fashion.
The inventory/shipping tracking systems for ecommerce are either clunky, or priced out of reach for Mom and Pop eBay sellers. I can't think of any CMSs that are particularly helpful for that vertical, either.
You could make a bunch of people happy by fixing any of those (okay, I may be the only person chafing for a good surface subdivision tool for Android). Unfortunately, you're going to have a hard time making real money from any of those niches.