1. Compiling too early and too often.
Now you are spending your time working on fixing formatting and such that will change as soon as you add more text anyway and all the floats will move anyway.
2. Using floats when you don't have to. Floats are the devil, yet all tutorials use them. Most of the time I know where I want the image, at least roughly, and didn't know for ages that I could just insert the image without wrapping it in a float, so I'd spend ages trying to get stuff to stop floating down multiple pages, or into the wrong section, etc.
Edit: As this is attracting a lot of debate: When I started every example shows you to insert an image in a float, so I thought you had to use float. Even a couple of months ago I thought you needed to put a table in a table environment; I'd still have to look up how to make a non-floating table. This is a problem as a lot of the time I need an image in a certain place. For example, if I'm typesetting a homework problem (Something I did a lot as an undergrad, and am now doing again when I'm writing questions) I can't have that image floating down into a different question; my TA or prof isn't going to go hunting for it, they are just going to dock me marks (And I don't want to confusing anyone working on my problem set!). Also as the h option is a joke they usually float far away. I don't mind if a figure floats a few paragraphs, but three pages away is bad: Same page as the discussion or facing page only. I should always be able to see both figure and text at the same time, unless there are exceptional circumstances or a very large text (I cut some slack when writing my thesis, as there were so many figures they wouldn't all fit on the same page as the text).
There are times floats are the right choice; I'm using one in the document I'm working on right now. However, they are just as often the wrong choice, and a lot of new users don't realize this.
3. Using outdated packages that some website gave you My group drives me crazy since they use very outdated packages that have been superseeded.
subfigure instead of
subfig, stuff like that.
4. Relying on GUI text editors instead of understanding what is happening People at work also drive me crazy with refusing to use anything but the compile button in WinEdt. If that doesn't automatically work, they refuse to use the package (Thus no biber, no biblatex, etc.
5. Using MikTeX That install package on compile thing NEVER works right. Hard drives are large, dammit.
Edit: Ok, not never. But it fails every time I've given my coworkers a new package to use, causing us to have to go into the interface and have it manually install.
6. Limiting yourself to what chemistry journals allow for everything I dream of one day being editor of a big chemistry journal, just so I can go to the AMS or APS and borrow their code. This has things like not using macros, not redefining things, not using excess packages (or in one case, not using packages AT ALL).
7. Not being willing to put in time understanding LaTeX The people I work with just want a solution now. They aren't willing to put in any time into making it work; they want something that works now, since the journal will reformat it anyway, so if it works and gets vaguely close to what they want, good enough.
8. Making DocumentNameV3.tex and forgetting to open the new PDF, then wondering why nothing you do is working Not that I've done this recently, after I was too lazy to set up a new code repository.