Max Woolf analyzes the effect of offering tips as incentives on the output quality of large language models (LLMs). He explores whether monetary and existential prompts influence GPT-4's adherence to constraints and text quality. The inconclusive results after extensive experiments with the ChatGPT API reveal the complexity of quantifying and improving AI performance with incentives. The tests ranged from constraining story lengths to evaluating text quality using a specially designed ranker.