even we have seen in the past how crypto stuff can fail (e.g. DES or MD5) from a pure protocol prospective, the key challenge in cryptography is key management.
Regarding AES, there was an almost independant worldwide challenge looking for the most secure algorithm for the next decades (which is quite challenging concerning computer processing power and new ideas, like not only using traditional CPU-cylces, also using other components like graphic card processing power).
Some people may disagree, but Bruce Schneir's Crypto Gram is the right start to talk about crypto philosophy ;-)
Regarding truecrypt (I'm also prefering container encryption for my sensitive data over full filesystem encrpytion)
Schneier on Security: TrueCrypt
Also concerning truecrypt attack vectors, read this:
Bootkit bypasses hard disk encryption - The H Security: News and Features
Get a trusted copy of truecrypt..USE COMPLEX PASSWORD and good entropy to generate container, don't save history or cache passwords! Choose the right algorithm combination, based on your need of performance+security (use truecrypts benchmark tool).
P.S.: even some people think our *N*S*A* (simple obfuscation ;-) friends still have endloss access to computing power & master minds - attacking an AES256 and higher, using SHA512 or higher and so on...the will likely use a much simpler way: social engineering ... and not waiting the next 2 years to decrypt your stuff ;-)