Though not speaking from experience it strikes me that having a prototype done with all the programming and game mechanisms worked out. Allows for better and quicker level design because you know how the enemies are going to react etc. and thus you can create more interesting levels that use the game mechanisms properly wehereas if you don't actually have the programming done you might find out that the things you planned don't work 'cos the programming doesn't work sa hoped etc. which would then require redesigning of levels.

Yea, gamedevelopment does not start out with creating an AK47 and
some Textures....
This might be ok for the pure hobby-development fun, not for a professional approach.

Milestones are important for me to keep things in order. If there would only be deadline, I'd do things in wrong order, and the whole project would be messed up. Also milestones give out the general idea how the hingyfoo should work, and you don't rush to things, leaving other functionalities behind and not up-to-date.

Deadlines, in other words "The time you get it ready, or you are kicked out". They make you keep working, if they are set in near future and you have hella lots to do. If they would be no deadlines, ye would be slacking and widening the development time, just like Sony.

Prototype computer for me is my own computer, the one I think will be the mid-level at the user-group. However, for development, you do need better than you need to run one.

Some poor running games are probably linked to
"high-end" computers of the developers.

It is hard to see framerate issues, if you develop on a machine
that is on the top-end of the targetgroup.
(the game is developed for the customers, not the pleasure of the developers)
The developers should at least have a low level machine permanently ready for
testing.
At least the person who assebles the game-content
(lead programmer) should work (have available)on a mid-range machine.

Will Wright, Game Designer extra-ordinaire, is the biggest advocate of prototypes that I know. At a recent conference, he said (and I'm paraphrasing here) that he will make 10 prototypes for every one feature that actually makes it into the game. These prototypes are little more than stick figures in 2d running around the screen, but like others have already said in this thread, you get a feel for whether a feature will be fun or not immediately. He then went on to say that for every 10 feature prototypes that make it into a game, one will actually go in with the code as it is. In other words, and this is important, code reuse is NOT a priority for prototyping according to Will Wright. Get it done quick and dirty and buggy, but get to the gameplay ASAP. Optimize and make it stable later.

fastlane: that system of him in EA is the cause EA sucks, I've seen so many games that have good gameplay, but crash every other second :|