Compression and LLMs are trying to solve the exact same problem: predicting what comes next.