The Predictive Power of Philosophy: Why You Can’t Ask a Gun to Read a Bedtime Story

Short summary

Understanding a tool's inherent nature—what it is and isn't—determines how safely and effectively you can use it. LLMs are non-deterministic statistical engines, not deterministic calculators; forcing them into rigid constraints damages their value. Pure functions and strict typing prevent the chaos of unpredictable global state and runtime mutations.

• •Tool limitations enable prediction and safety; misunderstanding them creates chaos
• •LLMs are non-deterministic by design; constraining them undermines their core capability
• •Global state and dynamic mutation in systems produce unpredictable behavior

Generated with AI, which can make mistakes.