After thinking a bit, here's some advice from me specifically:
- Deadlines are good. 7DRL, that sort of thing. To encourage finishing.
- Testers are good. If you can rope a friend or two into playing the game for you it helps a ton. I put a keystroke logging and playback system into ThiefRL and got play session recordings from my brother and a friend, and it really helped me find spots where the game was confusing.
- If making a game is your goal, always ask yourself what the simplest solution that could possibly work would be.
- Make sure that AI systems are focused on the player experience. If there is a complex simulation that isn't communicated well to the player, it just looks like random things happening, or worse, is never noticed at all.
- Making games is an endurance sport. It takes years of effort, each piece built on the ones that came before. A lot of the indie hits of recent years are 5-7 year projects, which is a bit of a slog.