Overall, the movement and mechanics here are solid (especially with configurable controls), but I found myself noticing problems with the levels themselves. As a puzzle designer myself, I hope to be able to communicate how I think about progression in puzzle games and why something felt a bit off about the levels.
First of all, quality of life is something that can and should be baked into the levels themselves. That means making it easy to execute not just the solution, but also any experimentation that takes place first. Most of these levels do this well, but I still feel like it's important to remember going forward. Level 3 in particular I wish was a little smaller, and had a "catch" in the ceiling so that the rising block wouldn't squish you.
Also, being able to replay previous levels is more important for puzzle games than other genres, because you might want to process what just happened and learn from it. I would say that, in this demo, backtracking is possible but discouraged. The levels are indistinguishable from the hub, and a completed level will be missing its key entirely. These have easy fixes though, like making the (strangely hidden) level names visible from the hub, and leaving a grayed-out or dotted-line key in finished levels.
Being able to learn brings me to my last point; while the levels have a progression of difficulty, they don't have a clean progression of knowledge. Having unique eurekas is fine, but when none of the ideas are built upon and everything is different, there is no sense of learning how to solve the game's puzzles. This leaves them feeling disconnected, and effectively, having the same difficulty for a new player, who goes in knowing none of the ideas used for solutions. Going further, when you reach a puzzle that requires you to consider the consequences of a fundamental mechanic, such as solid objects being stopped by other solid objects, the player should already be familiar with that mechanic. However, I found myself discovering and applying a mechanic in the same level, to the point where the puzzles could be considered "unfair." Don't get me wrong, figuring out new mechanics is enjoyable, but most levels turned into poking and prodding, trial and error, instead of logical thinking and planning ahead.
On the other hand, I feel that a few of the levels just aren't good teachers. Level 2 can be solved accidentally, risking a player missing the finer points (the magnet dropping from the rotating door, and the magnet being able to fall through platforms). Level 4 is simply unclear; I found two different solutions, and I'm not sure if either of them are intended, nor did they teach me useful techniques or ways of reasoning about the game elements. "Cheese," as unintended solutions are known, are fine when they are much harder to find than the "real" solution (making for some interesting speedruns), but cheese can undermine a puzzle and its lessons when it is a lower mental hurdle than the intended solution.
While I have pointed out a lot of design decisions that you might consider changing, there is still a lot to like, and I would encourage you to keep building on what you've learned about game design. Level 5 was probably my favorite, being a pure puzzle with a non-obvious solution, and the movement feels great, making a great environment for poking around the puzzles (although throwing the magnet might need a touch-up, I didn't realize that the mouse controlled throw direction at all). Still, I encourage you to find a way of thinking about how players approach puzzles, even if it doesn't match my own puzzle design philosophy outlined above, because player knowledge becomes more important as you add more complex levels.