Free course · 11 lessons · about 3 h 15 min

Prompt well. Save tokens. Build agents.

Short lessons in plain English with exercises. Start with prompting and tokens, then build, save and safely connect your first AI agent.

🔒 Free, with a quick email sign-up

Unlock the free course

The course is 100% free. We ask for your email so we can send you new lessons, agents and prompts, and so you can opt out any time. After you sign up you'll land right back in the course.

Already on the list on another device?

You're in. Thanks for joining! Your progress is saved in this browser only.

The lessons

Take them in order, or jump to the one you need.

Part 1 · Foundations

  1. What Is an AI Agent?

    Chatbot vs agent, the Goal → Plan → Act → Check loop, and why your agent is a robot intern, not a genius.

    ~10 min

  2. How to Prompt Well: The Intern Brief

    The six-part structure (role, context, task, format, examples, constraints), with before/after prompts and exercises.

    ~20 min

  3. Iterate, Verify, and Keep What Works

    Good prompts are rewritten, not born. A 5-step iteration loop, how to fact-check an AI answer, and how to build your own prompt log.

    ~15 min

Part 2 · Tokens

  1. Tokens and Context Windows, Explained

    What tokens are, how to estimate them, what a context window holds, and why AI “forgets” in long chats.

    ~15 min

  2. Token Optimization: Shorter, Cheaper, Sharper

    Eight habits to keep prompts short, reuse and summarize context, and understand cost and usage limits, without making answers worse.

    ~20 min

Part 3 · Agents

  1. What's Inside an Agent: The Agent Brief

    Design an agent on one page before you build anything: goal, instructions, tools, context, limits, and a truth test.

    ~15 min

  2. Build Your First Agent: The $40 Dinner Planner

    Build, run, and honestly test a Dinner Planner agent in any free chat. No coding, no purchases.

    ~25 min

  3. Make It Reusable: Projects and Skills

    Save your agent so you stop retyping it, using a ChatGPT Project, a Claude Project, or a Gemini skill. And why we don't build on custom GPTs or Gems.

    ~20 min

  4. The Power of 1%: Compounding, Honestly

    The maths of getting 1% better every day, the honest limits of the idea, and two agents (Daily 1% Review and the 5-Minute Task Killer) to practise it.

    ~15 min

  5. Integrate It Into Everyday Tools

    Connect your first two-step automation (form or email → spreadsheet) on a free plan, with fake data first and safe limits.

    ~25 min

  6. Stay Safe and Keep Going: The Agent Safety Card

    Prompt injection in plain English, what never to paste, and a personal Agent Safety Card you'll actually use.

    ~15 min

What you'll be able to do: write a clear prompt in two minutes, estimate and reduce token use, write an Agent Brief, build and save a Dinner Planner agent, set up a safe two-step automation, and keep an Agent Safety Card.