A few months ago I started experimenting on a car controller. I got really curious about the many ways you could go about implementing car controls, from going full realism to just a sphere disguised as a car. I've been working on it on and off for a few weeks, sharing GIFs on Twitter and having a good time.
Then, other projects took priority, and thus it was time to revive the car controller and turn it into a game for #FragmentJam! Here are a few things I learned:

FIRST: Build a level.
I spent weeks tweaking different properties of my car, changed code, improved stuff and kept going back to earlier versions because the "feel" was better. The thing I ignored while working on the code: where do I want to drive my car? Will it run along walls? How realistic does it need to be if I want to make an arcadey game? Will my level have a lot of verticality? How much time will the car spend in the air?
These are all questions I left for my future me to answer. To be honest, not knowing how my level would look made me waste a lot of time. So at one point I started building, blocking out a level, and making sure the controls for the car were suited for traversing the space.

SECOND: Don't build a level. (when short on time)
I started too late (due to reasons already mentioned), which made me skip actually planning the level. I just started throwing blocks in there, treating the game like the test-environment I built for testing car controls. Also, since I am by no means a routined level designer (pretty much only did levels for one of my games before), I did not manage to build a really interesting map, merely a map that you can play on. This is fine by me for the sake of this jam, but it made me realise that I could have gone with procedurally generated content instead, which at least provides replay value by changing up the map, even if it is not the most interesting gameplay space.

So now with the jam coming to a close, I can safely put this game down for a while, and who knows, maybe I'll return to it someday.