command vocabulary for AI work
Terms that make language models easier to operate.
LLM Term reads AI vocabulary as working language, not decoration. The desk examines how phrases like grounding, retrieval, context, hallucination, latency, confidence, tool use, and evaluation shape real decisions in prompts, product docs, governance notes, and search-visible explainers.
term
grounding
Use only when a model is tied to retrievable source material.
term
context window
Treat as a budgeted working surface, not a memory promise.
term
evaluation set
Name the task, sample shape, scoring rule, and failure tolerance.
term
tool call
Separate the model decision from the external action it requests.
terminal brief
Every entry has to survive use.
We ask what the word does inside a prompt, inside a product requirement, inside an evaluation note, and inside a public article. If the answer is vague, the term is flagged. If the term clarifies a boundary, it becomes part of the operating vocabulary. This makes the site useful for writers, builders, reviewers, and teams that need a shared language before they argue about model quality.
01
A term is useful only when it tells a writer what to include, what to omit, and what evidence would make the claim stronger.
02
Prompt language changes behavior when it names the job, the boundary, the artifact, and the review standard in the same breath.
03
Answer engines reward pages that make their definitions extractable, sourced by context, and easy to quote without losing meaning.
The home page stands on its own.
Even without fresh article imports, LLM Term gives visitors a clear view of the editorial promise: sharper definitions, stronger prompt language, and concept pages that answer engines can parse without guessing. The archive exists for crawlers; the public experience is this working desk.