Prompt 3 - Information Weighting
Prompt 4 - Systematic Context Optimization
Yesterday you learned to think before writing and organize information in layers.
Today we get deeper into something most people never consider: every word you write has a cost, and understanding that cost changes everything about how you craft prompts.
You've probably experienced this... you write what seems like a perfect prompt. Clear thinking, great context architecture, specific instructions. But the AI gives you a response that feels off, like it forgot important details halfway through.
Here's what's actually happening.
AI doesn't have unlimited memory. It works within something called a "context window" - think of it as the AI's working memory for a single conversation. Every word you write, every word it responds with, takes up space in that window.
When the window gets full, the AI starts "forgetting" earlier information to make room for new information. So that crucial context you carefully placed at the beginning? It might get pushed out of memory by the time the AI reaches your actual request.
Most people don't know this is happening. They just think the AI got confused or wasn't paying attention.
Here's something that might surprise you: AI doesn't count words the way you do. It counts "tokens."
A token might be a whole word, part of a word, or even just punctuation. "Engineering" might be one token, or it might be split into "Engin" and "eering" depending on how common the word is in the training data.
Why does this matter? Because you're paying for tokens (if you're using paid AI), and more importantly, you're working within token limits that affect how the AI processes your request.
The average context window is somewhere between 8,000 to 1,000,000 tokens, depending on which AI you're using. That sounds like a lot until you realize that a detailed business prompt with examples might use 2,000 tokens before you even get to the main request.
Advanced prompt engineers design token-efficient architectures that pack maximum intelligence into minimum space... but that's way beyond foundation level.
For now, you need to understand the basic economics. You also need to understand when you can use token-heavy prompts and when not to. In this course, I’m using long prompts that will cost a lot of tokens, here’s why…