The biggest problem with the "killer app" idea (beside conceiving the thing in the first place) is competing with the mega-resources of the large software houses.
I don't know how many times I have had to explain this to people who want some custom programming done. Within a realistic budget you just are not going to get the bells and whistles of a commercial word processor, major web browser, or what have you. It's tough enough given a small feature set to have a program that is solid, has a decent user interface, useful online help, written documentation, etc.
I'm doing some custom embedded development work right now for example. Nothing especially innovative, just making use of modern technology to handle an old task a lot more cheaply. I am spending a tremendous amount of time on research into things I would have assumed were well known and well documented:
[ul][li]quirks of various embedded OSs[/li]
[li]even bigger quirks of embedded developer libraries and prebuilt applets[/li]
[li]odd characteristics of various flash memory devices[/li]
[li]unexpected cost trade-offs[/li]
[li]trying to find matches between embedded OSs and hardware devices (driver support and so on)[/li]
[li]frustrating licensing restrictions[/li]
[li]the general poor quality of open-source code (though some is excellent)[/li][/ul]And the list goes on...
Getting the main application converted to portable code took a chunk of time. Separating out the user-interface logic took another (there are few true portable UI libraries even in the Java world - and I'm not using Java).
I hate to think how many printed pages my lab notes would probably fill. Staving off feature-creep while I do all that other stuff is a huge challenge for this one-man band. The only saving grace is I get to keep everything I learn and develop except the final hardware/software design and documentation. That's a rare thing these days. But I digress...
My point is that a "killer app" (or killer hardware/software product) can be a ton of work. Then you have to slick it up for mass-market consumption after you bullet-proof it to avoid going broke on returns or support calls. At least my project above doesn't need market appeal, the specs are (yeah right) in stone.
