Use relevant technical terms

Using relevant technical terms means writing prompts in the precise vocabulary of the domain you’re working in, instead of paraphrasing around it. Saying “make the function not block” is loose; saying “convert this to an async function using await instead of callbacks” activates exactly the right region of the model’s training. Terminology is a steering mechanism: the words you choose determine which patterns, idioms, and quality bars the model draws on.

This matters because models mirror the register of the prompt. Ask a question the way a beginner phrases it and you tend to get a beginner-level answer, complete with over-explained basics and simplified (sometimes wrong) advice. Ask it the way a practitioner would — naming the algorithm, the design pattern, the RFC, the API — and you pull the response toward expert-level content. In domain-heavy products like legal, medical, or fintech tools, correct terminology in the prompt also dramatically reduces ambiguity about what you’re actually asking.

Practically, you’ll write prompts that name things exactly: idempotent, memoization, optimistic locking, WCAG 2.2 AA rather than fuzzy descriptions of them. When you build prompts for users who don’t know the jargon, add a translation step in your pipeline — a first call that rewrites their plain-language request into precise domain terms before the real task runs. Precision in, precision out.