editorial method

Test the word where it will be used.

The method behind LLM Term is simple but strict: a definition is not finished until it works in multiple places. We test a term against prompt instructions, product requirements, evaluation rubrics, support explanations, and public concept pages. If the wording breaks in one of those places, it is revised before it becomes part of the desk.

Research grid comparing prompt words and model responses

Prompt fit

A term should make an instruction easier to execute. If it invites argument about tone, scope, or success criteria, the surrounding sentence needs a more explicit artifact and boundary.

Review fit

Model review language needs observable evidence. We prefer phrases that let a reviewer point to the missing source, unsupported inference, format drift, or task mismatch instead of writing vague complaints.

Publication fit

Public pages should be readable by people and extractable by machines. That means clear headings, visible definitions, canonical framing, and enough context for a quote to retain its intended meaning.