Write down what your game is not, or small yeses keep moving the line

"It is a small thing, one of them will not hurt" can be harmless on its own, and that single feature usually does not break the game. The problem is that the yes widens the line a little, and the next idea gets weighed against the line where it now sits. An idea that would have been turned down at the start now looks like it fits, and the game spreading this wide is what breaks immersion.

You can see the biggest examples of this in live service games. WoW is not in the tone it started in, a lot of what is in the game now has gone Disney, and Destiny was a hardcore FPS until the Edge of Fate expansion turned the player into a ball and had them solving puzzles. What players say at that point is "sure, but what has this got to do with anything?" Tim Cain, who made Fallout, calls it creeping normality. FFXIV is the counter-example: because the game was treated as a theme park from the very beginning, a huge number of different features can work inside one body.

In the first Fallout, what you called a robot was Mr Handy and Robobrain, clumsy machines on treads or hanging in the air on thrusters that it would never occur to you might be intelligent. Somebody on the team who loved Terminator kept trying to slip a slim endoskeleton robot in, Cain turned it down every time, and in the end he had to say this is not Fallout. In Fallout 4 you run into Nick the detective robot the moment you walk into Diamond City, and behind him come synths you cannot tell apart from people. The endoskeleton robot is normal in Fallout now, and the precedent is ready for anyone who wants something more modern in the next game. Keep going like that and one day the game will not look like Fallout and it will lose the appeal that made it.

The design doc written for Arcanum 2 had a stone in it that combined magic and technology in a single item. Cain objected to that too, because all of Arcanum's tension comes from those two never combining. The elves are against technology, the dwarves love technology but keep it to themselves, the humans want to spread both everywhere. The moment that stone goes in, a magic gun becomes possible and so does magic that imitates technology, and the thing that makes the game what it is gets closed off by one item.

After a few hours of play the player should know, or be able to guess, what can and cannot happen in this game. It would be strange if a flying saucer landed instead of a dragon in a high fantasy game like Skyrim, and strange if characters started casting spells in Deus Ex. So it is worth writing down the concepts that do not fit the game, and the ones you tried and did not like, as much as the ones that do. Because when you look back, if there is no written line, every idea that comes to mind can get into the game and the scope swells. Then there is the other habit, insisting on forcing a feature or an asset that sits badly in the game until it belongs. Even if the thing eventually becomes normal, it carries on sitting badly, because people are wired to match themes to each other and look for cohesion on a visceral level. And the effort spent purely on making it look right comes out of the time you would have spent making the game good.

You can see the same drift in Black State, a Turkish game that has not come out yet: the game has evolved into somewhere completely different from where it started, or at least it looks that way from the outside. Mechanics with no cohesion between them have been mashed together in the name of sci-fi, and what has come out is a game you understand nothing about at first glance. Unfortunately none of the material I have watched so far has given me an idea of what the game is, not the interviews and not the game's own footage. I hope it is very different once you play it.

None of this means close the door on ideas. Pushing up against the line is allowed, the place where you have to say no is past it.

The line you did not write down gets drawn by whichever idea arrived last.