You seem to have figured this out already: A new product is probably the easiest for them (because there's less ramp-up learning your code) and for you (because

You seem to have figured this out already: A new product is probably the easiest for them (because there's less ramp-up learning your code) and for you (because you don't have to deal with their crap code if you don't want to). Similar to this would be a simple orthogonal feature loosely coupled with your own code. First, sit down and define what sort of new product/feature you want, and determine how long it should take. Then hire enough interns that it'll get done within the time frame you need, and be generous. If it goes well, you'll have that many people you can easily hire later who won't need as much training, plus a new product on the side. If you don't hire enough people, you're just wasting everyone's time, because they'll be too frustrated and lost to do well, they won't want to come back, you won't want them back, and you won't even get that new feature you want. Career centers are good, as are online communities like github and hacker news. Of course, they're only an option if you can pay well enough to have people travel to you, or are willing to consider remote work (this usually is very difficult, try not to need it).