I recently watched a lot of Youtube videos about Game Development, and one idea that was mentioned many times was developing a prototype before you start the developing the actual game.
"Prototype" means a really, really rough version of the game with no real graphics, basically "triangle vs squares" or something like that.  

You are supposed to test your game idea if its actually fun. If it is, you are supposed to scrap your prototype and start over to develop the real game. 

I'm wondering if this makes any actual sense.

First problem, playing "triangles vs. squares" might be vaguely fun - but for most games, having decent graphics, music, good sound and great controls are kinda essential to have any fun at all. 

Second problem, it feels unnecessary for projects in established genres - if you are for example developing a jump & run or a regular RTS, do you really need to test if the idea is fun? We already know these genres are fun, at least to the right kind of player.
I guess if you have a super unusal idea, a prototype would make way more sense.

Thoughts anyone?

for most games, having decent graphics, music, good sound and great controls are kinda essential to have any fun at all

The simple reason prototypes don't need those stuffs is they need to be made quick. Omitting those stuffs is one best way to do it.

Also have you played any game with decent graphics, music, good sound and great controls but horrible gameplay?

 if you are for example developing a jump & run or a regular RTS, do you really need to test if the idea is fun? We already know these genres are fun

Sure, those genres are fun but the real question is : is your take of the genre fun? Ex: if you were to make an RTS, are you sure it would be fun?

Back to my point above, aside of being quick, removing graphics, music and sounds would make it easier to focus on the gameplay. There are many aspects to set up properly to make it fun.

I'm wondering if this makes any actual sense.

Context matters. A lot.

You are not saying who wants to build a game, what type of game, what the scope is, if there are investors, if it is a hobby game, established franchise, experienced developers and so on. All these and a lot more will change the context.

If you invent something new, you need a prototype as a proof of concept, before you commit resources. Lest you find out that your free energy device will not ever work. A thing you might have found out by applying common sense and laws of physics. But creating software for people to have fun with is not exactly hard physics.

So you might want to recheck the target audience of those videos.

Also, some people just call their pre-alpha version a "prototype". Or do not bother at all to use classic alpha beta release naming conventions.

/games/prototypes

Yes, it makes sense to develop a prototype. So you can have something to play and test your ideas early on.

BUT, you don't have to scrap your prototype. You can replace its simple graphics with polished art and add your sound tracks to turn your prototype into a final game. Ideally, you want to build a modular prototype where you can add or remove elements you want or don't want in the final game.

Yes, it is always recommended to start from there. Starting from the art, the 'end look' is an awful idea. Art is more or less easy to produce (or get using 3rd party assets) but also is time consuming. 

It's recommended to start from all the 'systems' (character controller, enemies AI, triggers, etc. Specially if you work in engines like Unity, GODOT or Unreal, and similar.) because that way you can see if an idea you had is easy to do, makes sense, is fun, etc (and that way you don't spend weeks or months working in a enemy or a level that will have to be discarded).

 It's not necessarily to  use only primitives (cubes and triangles), if you need you  can use temporary simple art (placeholders) to get an idea if the game feel and looks may be remotely similar to the what you have in your head.

Fun may be more or less subjective, but when starting the game, the most important part is make sure the ideas you have can be executed.

Well, I agree with you on some points : Some genre may not need a real prototype to test the gameplay. I was thinking especially about RPG's (unless you innove in the field of combat system). But I guess in every genre you can add a special thing that makes your game special. RTS for example, you may need to test a lot to find the balance between macro and combats. 

Upon all, I think this is better to go with a prototype, just like a painter would go with a sketch before lining and colouring his drawing. It could give you a better overview of what time it will take for instance. Also it is sometimes easier to go step by step, little by little, creating your game, especially if you're very demanding with yourself. I know by experience I work quicker when I say to myself "Come on, just write, don't think to much, it's just a raw shot. You'll do better later". If i wanna try too perfect on the first shot it kinda paralyses me cuz i never feel like it's good. On the other hand when i go back on editing my prototype I can find the results better than I could have imagined. All I need to do then is just polish a little, and process again.