Exactly what I was thinking.

and part 2:

Most of these are done with Python using ImageDraw from the Pillow library.
Currently, I'm not publishing the source code. Honestly, it's nothing pretty to look at (It just needs to work). But I might publish all the source code on my GitHub later, when I'm done and make the final compilation video.

Nothing new from the problems themselves, but I have acquired new skills in doing visualizations in Python and Blender - super fun to see algorithms come to life!

Maybe it's the way you're detecting loops? I used a set to store 3-value tuples (x, y, orientation). Every time the guard moves, you add to the set. If a tuple is already in the set, that means the guard has already been on that exact location facing the same way and is therefore stuck in a loop.

Even with that optimization, your solution would still take minutes to complete, so something big could probably be done more efficiently. My Python code took 12 seconds to complete.

First of all, I only checked the coordinates the guard actually visits. Placing a new obstacle that isn't in the guard's way does nothing. That already eliminates around 70% of the possibilities.

Wow! This is cool!