In some grad schools you're assigned projects, but at others you need to continually figure out even what problems need to be solved and which avenues of attack

In some grad schools you're assigned projects, but at others you need to continually figure out even what problems need to be solved and which avenues of attack would likely yield the best results. This is much like what startups do, and in fact figuring out the important problems/attack strategies is much harder in my experience than performing the experiments themselves. I don't see how this is "hand holding", but perhaps the grad schools with assigned projects feel more like a normal job.