I had never had a headache in my life. However, since mid-February, I have often been annoyed by a persistent headache. Today, I found the cause of this symptom: it is the volatile response of LLMs when I prompt them or use AI-powered search.

Every output from an LLM changes without cause, even when I use the exact same script, phrase, or term. As a computer scientist, I was educated to believe that science is about finding rules in nature, and engineering is a method for implementing those rules in practice. Rules and methods are expected to provide something persistent and reliable that everyone can depend on. 

But LLM responses change every time you ask for something. It is absolutely getting on my nerves. I truly hate the fundamental technology underlying LLMs.