Focus on the UI and UX of software that runs on GNU/Linux to promote mass adoption. Great, reliable programs don't help when the casual layman can't use them (no UI/all CLI) and things that look better often sell better than things that function better.
More users (and thus a larger market target for programmers) may mean more talent to improve GNU/Linux.
I'm talking more about form-function and a GUI interface being provided as an alternative to a CLI. Don't have 500 options buried under the "Options" menu. Be a bit biased on 'trivial' details to lower the amount of information a user needs to know/remember. Organize things in a reasonable manner.
I've seen a music player where the play button was on the bottom left corner but if you wanted to skip the track or pause - the buttons were on the right. If you wanted to load a new directory to play music from you had to use "black magic" (read: hotkeys) of "ctrl+u" to open a prompt that asked for the directory to read from. As you might imagine - the UX was terrible.
The biggest problem with the Linux ecosystem is that windowing systems and X-Window (and clones) in particular are an inefficient design. They're given up performance, increased complexity, and made writing applications more difficult than they need to be. The only trade-off people keep mentioning is "you can now run it over the network!!!" which is something few users do, and VNC (and similar remote desktop solutions) remain very popular (since remote windowing is complex and problem prone).
Essentially I'd design the UI/GUI experience more akin to Android (or failing that at least monolithic like MS Windows, but without the Win32 problems). In 2015, it may be HTML5/JavaScript/CSS based like Windows Store/iOS/Android apps, as that seems to be the direction the market is heading anyway.
The fragmentation problem is really bad. I'd encourage people to focus on one core upstream distribution and then fork it downstream closer to the specific point of customization rather than creating over eight different upstreams. Alternatives are good but not when they're all basically the same. I'd only encourage root-level distro forks when people really want to do something fundamentally different.
The kernel is mostly fine. I'd probably have merged OpenVZ years ago, which would have given us containers (and pretty secure ones to boot) much sooner.
It's a hackers dream, if you're in college and you want all code to be available for you. But when you get in the real world - you realize your time spent programming is more valuable. Maybe you don't want to grant the world the rights to your creativity.
Maybe the GPL can be credited for the success of the kernel's evolving. But there is little evidence that GPL brings serious players in - commercially - GPL wasn't invented by economists - it was invented by an academic.
I'm surprised to see Valve using Linux and not BSD. BSD is even simpler from a developer stand point. I love debian - but as I get more serious about my skills - my investment is going full boar into permissive tools.d
I'm free to borrow code from these projects 10, 20 years down the line for use at work, to add value, no questions asked.