However, my bigger concern is that working out the requirements is being done by someone other than the developer that's going to be working on the project. If you're trying to get something done on a budget, fewer people in the mix is better.
Furthermore, the conversation about product requirements needs to be a two way discussion. Developers are often in a unique position to guide a product owner into what's easy to implement and provides the biggest "bang for your buck." Their early insight can radically shape the product and is not based on what the customer thinks they need, but what the customer communicates their goals are. This is where a consultant can provide serious ROI to a product owner.
So many things can go sideways, and at a 4% take on $3,500 projects, as loosely defined as that is, there is no way this can scale.
Sure there is, limit to what time they spend working on getting projects ready, if they don't meet a certain standard, drop them and give quick 5 second tips to get them back in the queue. If they don't produce a reasonable spec in enough attempts just drop them as a user.
Even the AP had a work around for scaling this further, run at a loss for some time while you identify patterns of failure, produce resources for the user to solve their own problems.
For $140 I could easily spend 10-20 minutes looking at a rough spec and tell you if it is workable for 3500, I'd even be able to put in a few slash points on which features would likely need to be cut.
For some perspective, $3500 @ $50/hour = 70 hours. In other words, 1-3 weeks of dev time.
The types of developers you want creating your prototypes are those that can justify charging $200+ per hour because they have past business/user experience to truly understand what you want to achieve from your project as opposed to what you've specified.
Also as another has pointed out, justifying 200/hour and beating the competition is very difficult in a world wide market. Those in sv may be able to push that high but I can assure you there is many many developers with the same skill set living in much cheaper economic climates where 50/hour is considered much higher than 200/hour is in sv.
Since the developer's location is becoming irrelevant I guess that's a reality for developers in general.
For example, trashbin and "i want an ebay clone" elance style requests, and filter out the developers who claim to be able to do it for $500 in 12 days.
That alone would cut down most of the crap you see on sites like elance.
As for their fears (dsrguru posted) about people going out of band to work around 4% for a measly $140 I don't see how that would really come into it as a reason for doing the filtering. $140 is nothing to a developer for a guaranteed payment on completion, its also against the interests of the buyer as they may end up with subpar quality and an angry developer chasing them down for work they didn't really do to a reliable standard.
Sure, there will likely be a few who try to work around it, but really they aren't the target market user to begin with (the i want ebay for 500$!!? types)