One of the adivce I always give to newcomers is start early

Well, you're limited to how much you can actually do, but here's some things that you can do to be successful:

The final theme voting round is coming up. As soon as it does, vote for your favorite categories, and brainstorm on as many ideas as you can that fit into those. It's extra helpful if you can come up with single ideas that meet multiple themes:

i.e I have an idea for a game that can follow the themes of "On/Off", "Automation", "Switching Perspectives", "You are the power source", "Control the Environment", "Delivery", "Missing Parts", "Making Connections", and "Out of Order", It won't be the exact game no matter what theme is chosen, but so many core mechanics are shared that I at least have a good starting point on my design.

For a previous game of mine, I already had a desire to make a board game with an AI-controlled opponent. When the theme "Keep It Alive" was chosen, my design was tweaked to include an egg you had to protect.

I recommend starting a text file and fill it with as many ideas as you can (and keep this file for future jams, you might want to use some of those ideas later!).

This much at least you can do during both the Jam and Compo, but there's even more you can do with the Jam:

Per the rules, you're allowed to start with any base code you have. You could start working on mechanics (ie if you already know your game will likely be a platformer, put together some movement scripts). I personally don't like to do core mechanics programming ahead of time, but technically it is allowed.

What I do are get started on the basic quaility-of-life stuff. A stand-in title screen, audio manager, playerprefs manager, pause and quit manager, things that I feel any game I make will need, but will likely be lost to time. I even put together some pre-made assets (1,2) for anyone who'll be helped by them.

Even having those type of minor things ready to copy prevent wasted time.