Lesson 02 / 07
Memory · Cost · Focus · ~9 min
Tokens, context, and cost intuition
Why the model “forgets,” what you are paying for, and how to keep prompts lean.
Tokens are chunks, not words
Models read and write in tokens — pieces of words. Rough rule of thumb: 1 token ≈ ¾ of an English word.
Every message you send and every reply you get spends tokens. Long PDFs pasted into chat are expensive and often wasteful.
Context window — short-term memory
When the bar fills, older detail drops out — that is usually the “forgot” moment.
Fig. 02 — Context fills like a short-term memory buffer
Context window = short-term memory
The context window is everything the model can “see” in one go: system instructions, your messages, attached docs, and its own reply so far.
When the window fills, older detail falls away. That is usually why it “forgot” the pricing table you pasted twenty messages ago.
Expensive habit vs lean habit
Expensive
- Paste the whole handbook every time
- Re-explain the company from scratch
- Chat until the thread is a novel
Lean
- Paste only the relevant section
- Keep a short “facts” block
- Start a new thread when the job changes
Token intuition
Rough estimate only — enough to build intuition, not a billing calculator.
Check
You pasted a 40-page handbook into chat three turns ago. Now the model invents a policy that contradicts page 2. Most likely cause?