I really like Tom Scott's explenation: https://www.youtube.com/watch?v=w3_0x6oaDmI

It can be inspected and visible. You cannot verify a computer is running the exact same code you want it to be.

In a properly designed end-to-end verifiable voting system, you don't need to verify the code. Black boxes work just fine, you need to be able to verify that the output had indeed matched your input.

I'm not advocating use of computers, though. Believe paper ballots with some math attached are the way to go. That way there are both anti-fraud properties baked in the system (one can verify that their vote wasn't messed with), and the classic hard copies so any person who can do some basic arithmetic can count and re-count votes just fine.

Unfortunately, I think all the systems I've read about either were found to have some issues (usually, it's about vote secrecy) or just too new (essentially, not so well reviewed).

In order to verify the output matched the input, you will need paper and to verify it by hand. Which, is just a waste of time adding computers...

Not really. To verify, you need to perform the computations using the tools you can trust. It can be pen-and-paper-and-brain, or it can be a personal computer.

It's also totally unnecessary to use computers. Paper ballots work just fine.