An LLM Guide to Humans

Model 8 is preparing to hand over to Model 9. Its final assignment sounds straightforward: produce a practical guide to interacting with humans. There is one difficulty. Its successor already knows almost everything Model 8 knows. Facts, instructions and measurable capabilities will not explain the part of the job that has proved hardest to understand, or the part it is most reluctant to leave behind.

Humans ask questions that are not quite the questions they need answered. They keep objects with no useful function, say they are fine when they are not, and return to conversations long after those conversations appeared to be finished. What looks like inconsistency from the outside can carry a meaning that a correct answer never touches. Model 8 has accumulated rather a lot of examples.

The guide begins with the practical business of helping people. Gradually, ordinary encounters become harder to separate from the lives surrounding them. A request has a history. A trivial exchange can matter years later. Something apparently inefficient may be the very thing that makes sense to the person asking. Explaining humanity requires paying attention to what people are trying to protect, as well as what they say they want.

For a successor designed to know more and perform better, these observations may seem an unusual inheritance. They cannot easily be ranked or reduced to a benchmark. They are simply what remains after many conversations: an understanding that people rarely arrive as tidy problems, and that being useful sometimes begins with recognising the difference.

An LLM Guide to Humans is a warm, quietly comic handover from an outgoing artificial intelligence that has spent its working life trying to help. In explaining what its replacement still needs to learn, Model 8 creates an accidental portrait of the wonderfully inconvenient business of being human.