a visual course

The Price
of a Prompt

(AI Cost Optimization)

Why AI bills explode — and the five levers that cut them 10× — in about half an hour.

A free, visual course on the economics of using AI — why you pay by the token, how chat history quietly multiplies cost, and the levers that tame the bill: right-sizing models, prompt caching, trimming, capping and batching.

in the real world

These levers decide whether an AI feature costs $50 or $5,000 a month — they're how chatbots, copilots, agents and RAG pipelines stay affordable at scale.

Start with the meter →

Part 1 — Where the money goes

1

You pay by the token

Every AI call is metered in tokens — and the tokens the model writes cost several times the ones you send.

2

The context tax

The model is stateless — every turn resends the whole conversation. Long chats quietly multiply the bill.

Part 2 — The big levers

3

Right-size the model

Frontier models cost 5–10× more per token than small ones. Most tasks don't need the frontier.

4

Cache what repeats

Your system prompt is identical on every call. Prompt caching re-reads it at ~10% of the price.

5

Trim, cap, batch

Shorter prompts, capped outputs, and a 50% discount for work that can wait until tonight.

Part 3 — Cost-smart in practice

6

Measure, then put it to work

Cost per request is a metric like latency — track it, tune one lever at a time, and see where it pays off.