reference guide

Your First Agents — in ChatGPT, Claude, or Gemini

← Back to the exerciseCockpit

Doing the exercise in a chat tool

This exercise works almost unchanged in an ordinary chat product — an agent here is a prompt, and every chat tool can run a prompt. The only real differences: there is no system-prompt box (your prompt goes in as the first message), and nothing is saved for you.

Set up your tool first

  1. 1
    Use a temporary chat, or a fresh conversation per agent.
    In ChatGPT: click the model name → "Temporary chat". In Gemini: menu → "Temporary chat". In Claude: a fresh chat is enough; turn off memory features in Settings if your account has them.
  2. 2
    Paste your driving prompt as the FIRST message of the conversation.
    That first message plays the role of the system prompt: it programs everything that follows.
  3. 3
    Keep your prompts in your course document.
    The chat tool won't save agents for you — every version of every prompt you like goes into the document, verbatim.

The starting skeleton

Paste this as the first message, then play the cook

user prompt
Guide me through cooking a dish of your choosing. First, ask me questions — one at a time — until you know what I have to work with (ingredients, equipment, time). Then give me ONE step at a time, and wait for me to tell you what happened before giving the next. Steps must be action-oriented: "Do ___ with ___." Adjust your next step based on what I report.

Everything else — Step 2's format addition, Step 3's process from your domain, and all three captures — is identical to the main exercise. One experiment the chat tool makes easy: try dropping the driving prompt in mid-conversation ("From now on: guide me through X, one step at a time…") and watch an ordinary conversation flip into a driven one on the spot.

Back to the exercise →