I think you pretty much hit the sweet spot there by making them not too big, not too small. As long as Problems are out, like unintentional confusion on what actually killed you, this game is great.
By the way, I forgot to mention about that blur effect (or whatever it was). Basically I liked that one, it came the game more dreamy feel which suited it well. However, There was downside to it too, which had to do with my machine being too slow. For at that candle place where I had maybe 3 FPS speed, it actually made things worse, especially when I was still thinking that I had died because of having lit the wrong candle, which belief by the way was forced even more by only one of many candles actually getting lit. But otherwise I liked that effect, it was good idea to try something like that, making it also set itself apart from the rest of first person perspective games.
Thanks for encouraging comments about writing reviews. Although, it is easy to be the best when you are only one, making me also the worst.
Regarding my game my biggest fear is that it might be confusing in its current state. Just like you said, it took you a while to figure it out. Afterwards, finally looking from outside without time pressure anymore, I really should have put more priority to having better instructions. On last day I should just have abandoned levels 3-5 completely and just make other things better, like instructions. Then if Had time left, start making them one level at a time. Now I made all the levels as first thing, and when i started testing them, they even failed in the end, creating actually extra confusion when they all open after completing level 2 without any explanation inside game as to why. Just two lines of text at start of game telling that, and push mousebutton to continue would have helped already, but didnt do even that.
Originally those command buttons were just for testing, meant to have picture or text or something there, but in the end decided to save time and leave them as they are, after all, color coding was actually somewhat telling already when you just figured their purpose out first.
This was my first try on AI, although in some sense I did do it last year too, but in different way. This was very good learning experience, and although I would now, knowing better, do things differently and better, I think it came out quite well in the end considering it was trying out for first time. There are naturally problems, like when you attack buildings, units actually move faster than when doing those other two things. This has to do with the pathfinding system I did (theres different method for buildings and the rest), but issue like that is a minor issue, and important part is that it anyway works.