Of course it applies to any software, but some programs are more vulnerable than others. For example, when you want to have cryptography in your program, you use an established library, not write the algorithms yourself, because those libraries were written with security in mind (i.e. have protections against different kinds of attacks, for example, side channel attacks, in addition to being implemented properly). The whole point is to minimize the surface of attack, so that your system is more secure. And one way of doing so is to not give root permissions to programs that don’t need it (such as text editors like nano).

Again, like I replied to the other comment, most of the programs you need root for are designed with security in mind and are inherently more secure and have less vulnerabilities than a non security focused program (that is not to say that it is impossible for a security program to have vulnerabilities -it certainly occurred before and keeps occurring- they just have a lot fewer). But even if you need root permissions for a non security focused program, you still shouldn’t let any program have it, the whole point is to minimize the surface of attack.

What do you mean get root itself with a modified su? A program that has been run as a user cannot just get root permissions, that’s called a privilege escalation attack and is a serious vulnerability in the kernel which gets fixed quickly when found.

Sure, but sudo is specifically designed with security in mind as a security program, whereas text editors are not (although I am more likely to trust vim than vscode). Running a malicious program as the user and not as root can help mitigate the impact it could do, even though it will still be able to do a lot as a user.

It’s probably to protect against any potential security vulnerabilities in the text editor program itself, not to protect you from yourself.

In the loop where you iterate over the positions, you translate the matrix model with the position correctly, but when you rotate it in the next line, you rotate the identity matrix, not the translated matrix, and put the result into model, which resets the position. And so all the cubes get rendered at the same place which results in it looking like only one cube is being rendered.