This is a great set of tips (by @jmechner ) and a process I’ve also followed naturally over the years. If you’re in game dev you need to embrace these. Ironically the only time I break these rules is when dealing with publishers 🤣 Let’s go through them one by one! (LONG POST) (1. Prototype and test key elements asap.)
It’s important to test things immediately via the shortest possible method to get stuff visceral and on screen. Too many devs get caught up building systems or in idealistic logic traps, but all that’s important is “is this gonna work?”

Memoirs of a heisei programmer: In 1998 I relocated to Tokyo, an intra-Sony transfer from America and took a 30% pay cut just to get my foot in the door on early PS2 development. (Within 6 months, @yosp sheepishly boosted my wage by 50%, the reason why is in this story I suppose) ... I was put in a ps2 r&d group headed up by an older chap called Hagiwara which was behind a numlock’d door for security... but there was no hardware or anything so the lock was irrelevant at that point. I asked him what I should research...

Reflections of a Heisei programmer: During Star Fox we only did a few late nighters, and when I say late night, I mean working until the wee hours of the morning preparing a build for the lot check at Uji or something like that... ...I enjoyed them thoroughly as that kind of thing works as a catalyst to bond the team and the game was exciting to work on, I don’t recommend them often but every once in a while they can be fun...

When StarFox 2 was canceled we went for a big sayonara meal and Miyamoto explained to me that they didn’t want to be perceived as inferior to the newly released PlayStation and Saturn consoles. But because the contract between Argonaut and Nintendo was over I... ...had to go back to the Uk, and Miyamoto lamented that he couldn’t take me on as an employee because Argonaut had instigated an anti-poach after @giles had left Argonaut to join Nintendo a few months earlier after getting married to...

When developing StarFox, using 640k dos machines, we started running out of memory when compiling our code about half way through the project - in those days machine code was compiled in one massive blob, so what did we decide do? We re-wrote the assembler in a week to write out linkable object files, but we didn’t have a linker, so what did we do?