Negligible function
Cryptography's formal way of saying "too small to matter": eventually smaller than one over every polynomial.
A function μ(n) is negligible if, for every polynomial p, it eventually drops below 1/p(n). 2−n is negligible; 1/n100 is not. The notion underpins the whole asymptotic school founded by Goldwasser and Micali in the early 1980s. A scheme is secure if every efficient adversary's advantage is negligible in the security parameter, and that is how both semantic security and IND-CPA are defined.
The weakness is that "eventually" never says when. A scheme can be provably secure and still broken at 128 bits. That is why practical proofs for modes of operation now use concrete security, stating the advantage as an explicit formula in the number of queries and the block size.