Principle 1: Give Direction

This is part of the AI Bootcamp. Click here to learn more.

Vocabulary

• Constraint — A limit you set in a prompt — length, tone, audience, what to avoid, what a good answer must include. Too few constraints is the usual reason AI output comes back generic.
• Context — Everything the AI currently knows about your situation — the conversation so far, uploaded files, and saved instructions.
• Large Language Model (LLM) — The kind of AI behind ChatGPT, Claude, Gemini, and Perplexity. It generates text by repeatedly predicting the most likely next word — which is why it always sounds confident, even when it's wrong.
• Model — The trained AI system itself. GPT-5 and Claude Sonnet are models; ChatGPT and Claude are the products built around them.
• Prompt — The message you send to an AI. Its quality largely determines the quality of what comes back.
• Role assignment / persona — Telling the AI who to be ("You are a senior data analyst") and how to communicate ("patient, jargon-free, no fluff"). A role sets expertise; a persona sets voice.
• Training data — The enormous body of text and code a model learned from. It shapes what the model is good at — and where its blind spots are.