The Breakout

A workshop you run yourself, minus the lanyards. Twelve short sessions, one wall, and whichever AI assistant you already use.

Each session teaches one idea, makes you try it on real work, and sends you home with something useful. About four hours in total, in bites of 10 to 25 minutes.

Why you’re here

You’ve used ChatGPT, Claude, Gemini or Copilot. The answers are fine. You have a nagging feeling there is a lot more in there, and you’re never quite sure when to believe it.

Another article about AI won’t fix that. Twelve short rounds of practice on your own work will.

What to expect

  • Twelve sessions in four breakouts. Every card shows its time, from 10 to 25 minutes.
  • Every session opens with one idea in plain language, then you try it.
  • You use real work of your own. Every card suggests examples in case you draw a blank.
  • Each session shows a finished example and tells you where it usually goes sideways.
  • You leave with something reusable and one line on your wall.
  • Nobody is grading. A red dot is a result, and usually the more interesting one.

What AI fluency means here

Four skills, known as the 4Ds. Every session trains one of them.

Delegation
Deciding what to hand to AI, what to keep, and how to split the work.
Description
Telling it what you want well enough that it can deliver: the goal, the context and how to behave.
Discernment
Judging what comes back. Is it right, is it useful, did it do what you asked?
Diligence
Owning the result. Checking before you share, and being straight about how it was made.

The 4Ds come from the AI Fluency Framework by Rick Dakan, Joseph Feller and Anthropic, shared under CC BY-NC-SA 4.0. The sessions on this site are original and are not part of that course.

Where to start

If you work in a plain chat window

Start at session 1. Sessions 1 to 4 need nothing else. Sessions 5 to 7 use features most chat apps now include: web search, file upload and saved instructions. Sessions 8 and 9 come with a chat-window version. Read “Words you’ll see” first.

If you already use an agent with tools

Start at session 1 anyway. The early ones go quickly, and what you keep from them feeds the later ones.

If you’re bringing a team

There is a facilitator guide under For teams: four meetings, one breakout each.

Words you’ll see

Chat window
You type, it answers. It can only write back.
Agent
The same kind of model, given tools and allowed to take several steps on its own before reporting back.
Tools
What lets it act: search the web, read files, run code, use your apps.
Context
Everything the model can see right now: your message, files, earlier turns and saved instructions. It knows nothing else about you.
Saved instructions
Standing context it loads every time. Apps call these projects, custom instructions, custom assistants, memory or skills.
Fresh session
A new conversation with no history. Useful because it carries none of the last one’s assumptions.
Subagent
A second agent that the first one hands part of the job to.

House rules

  • Use real work. Fake examples teach fake lessons.
  • Go in order. Later sessions borrow from earlier ones.
  • Red dots are welcome here.
  • What goes on the wall stays on the wall. Nothing leaves your browser.
  • Snacks encouraged.

Put your name on the wall

There is no account and no sign-up. Your name and notes stay in this browser and are never sent anywhere.

Share my wall

A picture of your wall: twelve numbered sheets with green and red dots, and one of your notes on a large sticky note.

To save the picture, press and hold it on a phone, or right-click it on a computer.