Here is my compo entry, Autoexec. The basic idea is a shmup game with a level-up system, but where you have to decide all the upgrades you'll get before starting.

The idea was inspired by Dungeons of Dredmor, a roguelike in which you choose several classes from a long list before you start. There are still some decisions to make when you level up, but most of your decisions happen at character creation. Dredmor also has a "no time to grind" mode, which makes the dungeon smaller and the level-ups happen thick and fast. So it ends up being a game of coming up with a plan, designing a character around that plan, testing it out in the dungeon, and refining it based on what did or did not work. I loved the idea of a game that distills that idea down as much as possible, by having all the skill choices made before the game, and then a very short gameplay sequence where you test out your choices. In this LD, a shmup seemed compatible with the theme, and I realised I could have a go at this idea, in the context of a shmup instead of a roguelike.

Challenges faced

Making a scrolling shmup. This turned out to be not too hard. Most of the coding was straightforward. The one thing that turned out a little interesting is the puzzle of how to make guns that fire faster than the game's frame rate. My solution was basically that the weapons have to work out, every frame, how many bullets they should have fired since the last frame, and then work backwards to figure out where to put them. It worked out fine, although, due to the rush, I ended up re-implementing some parts separately for each weapon. For this and other rush-induced reasons, the code is pretty high in technical debt.

Making a game with so much content. To make a satisfying shmup, the game would need multiple enemies, and to make the level-up system work, it would need a lot of skill options. So this idea ended up needing a lot of content. There are 5 basic enemies, 1 boss, 5 waves, parallax background layers, 16 skills, a skill selection menu, icons for all the skills, sound effects and music. So there was not just a lot of coding to do but also a lot of graphics and sound content. I did worse than usual in terms of sleep. I saved a little time by letting some skills share sound effects. In my initial brainstorm, I had a list of 14 skills, but I ended up adding, removing and redesigning skills based on whatever would be fastest to code up. Somehow I ended up with 16. I didn't get to give the music as much care as I would like, although it turned out OK.

Playtesting/balancing. I ended up not spending nearly enough time on playtesting and balancing. If I had, I would have made the earlier waves harder and probably shorter too, and tuned the XP cost per level-up. As it is, the early part of the game is a bit slow/boring, and it's too easy to get the first couple of level-ups. It gets better later in the game. (The boss ends up being the best part. If I make another shmup it might be all-bosses, a la Warning Forever.) I would also have made enemy bullets bigger. These are all the kinds of details that are important, but hard to get right when everything is such a rush.

All in all, I think it worked out alright. I'm satisfied that the basic idea is a good one, and I think the game turned out reasonably fun, despite some balance issues.