When you're coming up with ideas you should have a prompt. Some place to start. It's hard to narrow down an idea if one just wants to make something neat with no direction. So start with an idea. Space marine for example.
Next is to start narrowing down different ideas. You don't, and probably shouldn't, start and stop on a single idea. You need to explore ideas so you come up with something new. This requires a special kind of creativity. We all have experience with the common space marine archetype. Close shaven head, muscular, big armor/space suit combo. Lots of military-esque aesthetics. And it may be a good idea to start off there so you have a basis to move away from. But really, if you have time, you can think outside the box. Does it really have to be a space marine as we know it? Does it have to be human? Does it have to be the size of a human? Does it have to have guns?
Whenever you are conceptualizing there are things you want, things you don't want, and things you don't know you want. The latter is where you explore with thumbnails. It mainly involves adding shapes and removing shapes. In a digital program, using your brush and eraser. On paper you just start drawing thumbnails. At this point you don't go into a lot of detail. Just get the basic look and feel of each thumb right then move on.
Think of functionality to narrow down ideas. What environments will this thing be in? Who made it? What level of technology was involved? When was it made? What is each part used for? All these things will effect the look. An alien space marine in the year 4050 will look quite different from a human space marine from the year 2090.
Boundaries are great because it gives each idea an identity. It makes it unique. Imagine a space marine that never landed on a planet. It would need equipment that helps it move in little or zero gravity. It would look different from a space marine that fought on planets. So don't be afraid to give an idea a limitation. "This space marine fights under water only." As you have figured out, a prompt with little or no limitations is hard to conceptualize. It has no form, no identity. If a concept can be anything, then it has little value, it's not special. Things are defined by what they aren't as much as what they are.
"Astronomy offers an aesthetic indulgence not duplicated in any other field. This is not an academic or hypothetical attraction and should require no apologies, for the beauty to be found in the skies has been universally appreciated for unrecorded centuries."