about the desk
A vocabulary desk for the messy middle of AI work.
LLM Term exists because AI projects often fail before the model is tested: the team uses the same words for different things. One person says context and means user history. Another says grounding and means citations. A product note says confidence when it really needs evidence. The site turns those overloaded phrases into practical editorial objects.
The voice is deliberately compact, direct, and operator-minded. Entries are written for people who draft prompts, evaluate outputs, design documentation, and publish pages that answer engines may quote. The goal is not to sound academic or futuristic; it is to make the next instruction, review comment, or concept explainer less ambiguous.
What gets included
The desk favors terms that change decisions: how to write a prompt, how to diagnose a weak answer, how to describe a model limitation, or how to make an article extractable for search and answer systems. A phrase can be common and still be excluded if it does not help the reader do something more precisely.
What stays out
Hype terms, vendor-only slogans, and definitions that collapse into “it depends” are held back until they can be made operational. The desk is small by design, so each page has room to carry context.