Ok, this is a real geeky post.
In the evolution of computer operating systems, there has been a trend toward smaller pieces working together, rather than larger pieces. The old operating systems like OS/360 were monolithic: everything was in one piece. UNIX decoupled utilities and the user interface from the OS. Linux has a plug-in structure that allows kernel modules to be added as needed. More radical microkernel architectures have the OS only doing inter-process communication and some scheduling. Everything else is done by separate processes.
I'm starting to see the same sort of progress in other software systems. The Apache web server is very similar to a microkernel operating system. Everything (including handling http requests) is done by plug-ins. The Eclipse IDE is very much based on a plug-in system.
My idea is for a microkernel type Customer Relationship Management (CRM) system. Older systems like ACT or Goldmine were pretty much monolithic: they weren't easy to extend. The current generation of CRM systems like Salesforce or SugarCRM have a plug-in architecture to extend the base system with new functionality as needed. The problem is that the base functionality can't be replaced. When I use a modern CRM system it's fairly painful. They have email clients built in, but they're nowhere as nice as GMail. They have mailing list managers, but nowhere as nice as MailChimp. They have time tracking, but nowhere as comprehensive (note I didn't say 'nice') as Quickbooks. What we need is a CRM system that has no inherent functionality: all the functionality is provided via plugins. That way I could plug GMail, MailChimp, Quickbooks, Basecamp, whatever I like into the microkernel of the CRM system and come away with best-of-breed everything. The limiting factors are, of course, the APIs of the constitutent parts (Quickbooks is notably lacking here). Perhaps the base system would be in the cloud, talking to a Firefox chrome application on the desktop? I don't know. But I do know that almost all the pieces are there to make this happen. Should it be open-source? Probably. Would this attract VC money? Perhaps. Tell me what you think.