The Baker's Dozen · Workshop 1 · session plan
Twenty people, two sessions, two hours each. Bring a laptop. We send a short questionnaire before we meet so the examples come from your work, not ours.
A laptop, and a Google account you can sign in with. The hands-on coding block runs on Gemini: it's the easiest way in and its free tier is the most generous. A Claude or ChatGPT login helps for everything else. The questionnaire checks all of this.
Why this technology matters, how it behaves, and which tools are good at what. Half of it hands-on.
Write a skill of your own, build a browser extension, and leave with a 90-day plan started.
The whole workshop hangs off fifteen working principles. Each beat below tags the ones it touches.
The goal of the first two hours is comfort. Everyone leaves knowing what this technology is, why it behaves the way it does, and which tool to reach for.
By the break, nobody in the room is afraid of the blank chat box. By the end, everyone has run real experiments with prompts, tokens, context and slop on their own laptop.
Relax, it's not here for your job Discuss
AI is a transformative technology, and it works best with a person in the loop. Teach it the routine parts of your job and spend the freed-up time on the parts that need you. The path the workshop climbs: Quoter to Curator to Author to Orchestrator.
Principles 01 · 02
How it actually works Teach
A language model in plain words: it predicts the next word, nothing more. That one fact explains most of what it gets right and almost everything it gets wrong.
Principles 03
Prompts, tokens, context and slop Hands-on
Four quick experiments, everyone typing. Build a strong prompt with the prompt builder. Watch a long chat lose the plot in the context window explorer, then learn when to start fresh. See what tokens cost and why shorter is often smarter. Finish by writing your own de-slop system prompt so output sounds like you, not like AI.
Principles 05 · 06 · 09 · 12
The lay of the land Teach
Who's good at what. The big chat models side by side on the same brief, live. Then the wider map: image, video and music tools, and services worth knowing like NotebookLM and the deep research modes in Claude and Gemini. How a prompt changes when the output is an image, a video, a post, audio or code.
Principles 04 · 11
Staying current without chasing Discuss
Tools will keep changing under us. What holds: the principles, your examples, your saved sessions. The habit that keeps you current is a monthly conversation, as a team, about what you're each using and what's changed.
Principles 01 · 13
Homework Homework
Between sessions: try the tools we compared, find the flavors that work for you, and talk as a team about which AI to standardise on. Revisit that conversation every month. Bring your pick to Session 2.
Principles 04 · 11
Session 1 was about using AI. This one is about putting it to work: skills that save your best methods, agents that do multi-step jobs, and a plan for the next 90 days.
Everyone writes a skill in the first hour and builds a small piece of software in the second, most for the first time. The plan at the end is made of both.
Recap Discuss
What stuck from Session 1, what you tried at home, and which AI the team is leaning towards. Questions welcome.
Skills: save your best methods Teach
A skill is a method you save once and reuse forever. What one looks like, where to find ready-made ones, and how a team shares a library of them.
Principles 06 · 14
Write your own skill Hands-on
Everyone picks one repeatable task from their own week and writes it as a skill with Claude's skill creator, installs it, and runs it on real work. See where it falls short, improve it, run it again. The skill you leave with feeds straight into the 90-day plan.
Principles 06 · 09 · 14
Agents, and the age of personal software Teach
An agent is AI that can take actions, not just answer. That unlocks something bigger: software you make for yourself, in an afternoon, without being a programmer. This is the biggest shift of the whole story.
Principles 01 · 08
Build a browser extension Hands-on
Everyone makes a small Chrome extension, live, from one prompt pasted into Google AI Studio and a zip loaded into Chrome. It won't be perfect on the first pass, and that's the lesson: look, say what to change, export again, three to five rounds.
Principles 07 · 08 · 09
What this looks like at full power Teach
A live tour of Claude Code: the projects it can carry, from motion graphics and video edits to websites, games and deep research reports. Not to learn today, just to see how far the same ideas stretch.
Principles 01
Your 90-day plan Discuss
Made in the room with the plan maker: four minutes, three lines, one hour a week on your calendar, one buddy, and everyone's card on a shared wall. Between day 60 and 90, swap skills: give yours to someone on the team and run someone else's. Identify the tasks you'll hand over, the processes worth turning into skills, and get the brand voice written down so everything AI touches sounds like you. Put system prompts and plugins in place so it all runs cheap.
Principles 03 · 10 · 12
Everyone needs a laptop, a Google account for Gemini, and a Claude or ChatGPT login. The questionnaire confirms this beforehand.
We pair confident people with cautious ones in the hands-on blocks. Both sides learn faster.
If a block runs long, we cut the talk, never the typing.
Covers both sessions and all the preparation, including the questionnaire review. The tools we use run on free plans, so there is nothing extra to buy.