I don't know how multi-quote works on these boards; if I did, I could pick and choose about a dozen principles from the above 8 posts to make my own list - I think every one of you has something I'd fully endorse, and you each also hit something I'd not want to use.
So, cribbing from the above posts in some cases:
--- character generation has to be simple. 15 minutes, tops, to get it table-playable; fine-tuned details can be filled in later.
--- character generation involves a moderate-to-high degree of randomness, players should be encouraged to come up with character concepts
after seeing what the dice say, not before.
--- character options (playable classes, species, etc.) need to be clearly laid out, along with the relative benefits and drawbacks of each.
--- gating some of those options behind roll requirements etc. to enforce rarity is acceptable
--- general principle: no benefit without an associated drawback, and vice-versa.
--- exceprience and-or advancement is a reward for what the character does, not the player
--- no metacurrencies (e.g. 5e's inspiration, luck, etc.) that allow a player to overrule or retry a die roll after the result of the first roll is known
--- the magic system should be powerful, but sometimes have risks attached both for the caster and anyone nearby (hat-tip
@pogre )
--- players should as much as possible approach any challenge by thinking what a person in that situation might plausibly do. (hat-tip
@Yora )
--- get modular. Taking a rule out [or changing or adding a rule] shouldn't cripple other parts of the game. (hat-tip
@GMMichael )
--- even if there's a best-playable or "sweet spot" range of character levels, make the design open-ended; i.e. no hard cap on level or progession.
--- design the zero end of the game as well as the hero end and let each table decide what part(s) of that they want to play.
--- PCs and NPCs are mechanically the same - a PC is generally representative of its class and species in the greater setting.
--- short-term balance between characters is a fool's errand, long-term balance is a worthy goal.
--- players have to know up front that bad things (including but by no means limited to character death) can and will happen to every character at some point, and the game has to be able to follow through on that threat.
--- the game needs to provide means to overcome or reverse those bad things, at some sort of major in-character cost or long-term penalty.
Well, that's a longer list than I expected.