This leads to two major issues: 1. The program will crash when the maximum context length is eventually reached. 2. Since we pay per token, this wastes money for no reason other than my own laziness.