Regarding Adium, I have seen some issues with it from time to time, but they tend to get fixed quickly. A good recent example was the wake from sleep crash I got 50% of the time in 0.86. I find the developers are pretty responsive to both feedback requests and bug reports for the most part. I had an odd issue when I was moving my system to work and back home with multimonitors. I typed up a description of the issue, and sometime in 0.8x the program did a better check to see if the contact list was visible. If it wasn't, it knocked it to a screen where it would be, solving my need to relaunch the program when moving.

Compare that to Proteus, my first IM program on the Mac. I went to the effort to make a XML translator to move my Trillian list over to Proteus, keeping my meta contacts and such. I used Proteus from version 2 to 4, and watched the development go from bug fixes to bug adding. They seemed to want to add some very advanced features, and would do so in release versions without much testing at all. Going from version 2 to 3, I was forced to manually readd all my ICQ contacts, because it finally went to server side contact support. No thought was put into migrating this automatically like Trillian did. Version 4 has stagnated as well with a promise of a 4.5 or something off on the horizon by the new developers for nearly a year now. All this in a paid product.

I'll admit, Adium is not perfect. And they have turned down some of my suggestions forcing me to use two IM applications at work (One for MSN/ICQ/AIM, one for Jabber chat rooms), but overall it is a lot better then the alternatives. I had considered GAIM via x11 once, but stopped looking at it when I realized GAIM lacks basic things like metacontacts.

Regarding Firefox, I only run it from time to time on Windows. I still find that even with 1.5, it's not an OS X application and stands out like a sore thumb. I wish Cameo was more an effort to make an OS X Firefox, complete with extensions support. But instead, it's just a OS X port of the Mozilla render.