Know what a model call costs and where it pays for itself.
Every class teaches one idea, asks you three questions on it and then puts it to work on a real company.
You pay per token. Output costs five times input. Ask for short answers.
Four numbers multiplied. Cache the instructions. Batch what can wait.
AI pays when unit cost falls and error rate holds. Count both.
Safe when the answer is in the text and someone checks. Needs a person when it isn't or the cost of wrong is high.
The engineer costs more than the model. Buy the 80%. Build the 20% that matters.
Baseline first. A metric you could miss. A risk with a number. That's the pitch.
Role. Context. Task. Format. And the facts in the prompt every time.
Read the ratio. Take the discounts. Date every price you plan on.
Automate. Assist. Leave. Decide it task by task with the person who does the job.
Baseline. Control. A person reading a sample every week. Forever.
Data. Vendor. Output. Each gets an owner and a check that runs.
One task. One team. One metric. Criteria before. A date to decide.
Figures. Model prices are from the provider's published pricing page read on Sep 7 2026 and will change. Every class that uses a price states the date. The validator flags any price older than 90 days for recheck.
| Source | Figure | Value | Date read |
|---|---|---|---|
| Anthropic pricing page | Claude Haiku 4.5 | $1 per million input tokens. $5 per million output | Sep 7 2026 |
| Anthropic pricing page | Claude Sonnet 5 | $2 per million input. $10 per million output | Sep 7 2026 |
| Anthropic pricing page | Claude Opus 5 | $5 per million input. $25 per million output | Sep 7 2026 |
| Anthropic pricing page | Batch API discount | 50% on input and output | Sep 7 2026 |
| Anthropic pricing page | Cache read | 0.1x base input price | Sep 7 2026 |
Ten minutes. No account. No card.
Take my first class