Bottom-up Dev

Bottom-Up dev

I write down the idea in a text file.

Then I write a bunch of "Wouldn't it be cool if _____" things about the idea.

Then brainstorm how the core features should work and throw those together. No comments, and shitty / fast code style for everything.

Once it actually works (because can't tell how exactly shit will need to be made until after it's been made) THEN I throw away that code. I re-implement it nicely the way it should be without the hacks and kludges. Then I go back to the "Wouldn't it be cool if" list and try one out, fast and dirty as hell. If it works, I might keep it. I make that a branch in my repo, and then revert and try something else (unless things depend on each other). Then I pick the most interesting ideas from those and merge them together. Once they work, I throw away those branches and make the nice beautiful commented and clean implementations.

All the while I'll be adding the odd "WIBCI ___?" ideas in that list and checking off ones that were shit. Very fucking important: Write down what was and wasn't cool as rationale under each idea so you don't fuck up and forget why and try to re-do some stupid shit that sounded good (again).

Get a prototype working with colored boxes. Then placeholder art. Iterate. Refine. Keep doing it.

Each dev day make a reasonable ToDo list of things you want to get done that day -- Then tear that shit in half, because everything takes longer than expected.
http://en.wikipedia.org/wiki/Planning_fallacy

Do that half of the list. SLEEP. I've logged SLOC and checklist progress. Crunch is actually slower than resting well and working hard with the time you have.

If anyone ever tries to set a deadline: Laugh in their face (this is very important), then point them to the wiki link above (not as important, they're morons).

Note: The daily plan is far more important than the weekly, monthly or game plan. If you don't have a good plan for the day, you'll piss it away on the AGDG.