I just want to stick my oar in here.
This thread has some great ideas, but for goodness sake - keep it simple. Don't start off doing an accelerator and then add so many things it never gets finished. Asking on an Amiga forum is always risky when you want ideas, because you'll see what happened with NatAmi (and look what happened there) - Feature Creep City.
Set your sights on a simple thing to start. Do that - gain experience, not just of the hardware, but of the market. Then work on something else. Don't start work on an 060 card and get carried away designing USB, SATA, SCSI, Firewire and everything else.
Let's get the most simple thing done first - just make a simple board. Proof of concept. Then worry about making things more complicated. It worries me that the thread started talking about just manufacturing open source designs of other people, and now people are talking about designing super-fast CPU accelerators with modern RAM and USB and stuff.
Let's not try and run before we can walk, otherwise we'll all fall over.