You arrive at a broken transport tube. Whatever passed through it is gone. Below the break, carts move along a track.
The interesting moment comes before you touch a control. You look at the tube, look at the carts, and begin to reconstruct what happened.
That is one of the puzzle ideas being explored for Fathom's demo. It starts with a situation you can read, rather than an instruction telling you which objects to collect. The environment gives you evidence. You supply the connection.
Give tHe player a reason to tHink
A puzzle becomes easier to care about when you already want something on the other side of it. Recover an object. Reach a place you saw earlier. Get a machine to stop helping in quite the wrong way.
That intention should come first. It gives every lever, passage, and moving part a context. You can ask whether something might help you, rather than trying every available interaction because you have no better idea.
This doesn't mean every solution should be obvious. It means the difficulty should come from working out a relationship or discovering a possibility. Confusion about the objective is a different problem, and usually a less satisfying one.
Let an experiment teach you sometHing
An unsuccessful attempt can still be useful. If changing a control moves a mechanism somewhere unexpected, you have learned a little about the system. If nothing happens, the room needs to communicate enough for you to understand why.
That is especially important when the world contains magic and unfamiliar machinery. Their behavior needs to become readable through use. A strange device can surprise you once; after that, what you learned should matter.
I'm also exploring ways the party's abilities could contribute outside combat, including remote interactions, moving heavy objects, and spatial tools. These are possibilities to test, rather than a promised checklist. The question is whether each ability gives you an interesting way to approach a room.