LLMs aren't remotely like compilers or power tools

Other: AI and machine learning(blainehansen.me)view on HackerNews
Large Language ModelsAImachine learningdeterminismmetaphorsdelegationprincipal-agent problem.

Author: blainehansen

Date: 7/20/2026

Article Summary:
The author argues that Large Language Models (LLMs) are fundamentally dissimilar to humans and deterministic systems, and that metaphors comparing them to compilers or power tools are misleading.