← → navigate · space next · n presenter notes
The Baker's Dozen · Workshop 1 · Session 2

Cultivated AI

Grow, don't Chase.
01 · Welcome

Two sessions.
Two hours each.

01

How it works, and what's out there

Why it matters, how it behaves, which tool for what. Half of it hands-on.

today
02

Agents, skills and a plan

Write a skill of your own, build a browser extension, start your 90-day plan.

02 · Two sessions
Recap · show and tell

Homework review.

  • Two saved back-office prompts.
  • Five prompts written by AI, one per medium.
  • One image, five versions.
  • One cross-check across three machines.
  • The team's pick: which AI do we standardise on?
03 · Recap

The fifteen.

Still the whole game →
01Learn for the AI of tomorrow.
02Not technical. It's communication.
03Choose → Communicate → Evaluate.
04Pick the right AI for the task.
05Frame the conversation, not the prompt.
06Great examples beat great instructions.
07Ask AI to ask you questions.
08Ask AI for help with AI.
09Iteration is everything. 3 to 5 passes.
10Don't take the quick win.
11Never trust one machine. Cross-check.
12Always humanize. Deslop everything.
13Conviction. Art. Sense. Judgment.
14Your sessions become your SOP.
15Every task = Know · Do · Show.
04 · The fifteen
The path this workshop climbs

The climb, day two.

Quoter
asks, copies
→
Curator
picks the best
→
Author
directs the work · today
→
Orchestrator
runs the agents · a glimpse
Session 1: 14 Quoters, 16 Curators. Today: everyone builds. author by day 90
05 · The climb

Skills are your SOPs for AI.

A method you save once and reuse forever. A small document the AI reads before it acts. Hand it the way you do the job, and it just knows.

write it once, every future agent inherits it
06 · Skills
What a skill looks like · plain text

Four parts.

01 · Name

A name, and when to use it.

One line. If you can't say when, it isn't a skill yet.

02 · Raw material

What you'll paste in each time.

The transcript, the reviews, the messy sheet.

03 · Method

Your steps and rules, in order.

Include the nevers and the always.

04 · Example

One finished example you're proud of.

Real input, real output. This part does the heavy lifting.

Built with Claude's skill creator and installed under Skills. Claude reaches for it on its own. On Gemini or ChatGPT, the same four parts live in a Gem or a Project.

07 · Four parts
Finding · trusting · sharing

Skills, in practice.

08 · In practice
Laptops open →

Lab 03 · Write your own skill

talkingtomachines.xyz/tools/skills

copy the card · paste into claude.ai · answer its questions · install the zip

09 · Hands-on
So what is an agent, really?

An agent is just this.

LLM + goal + loop + tools
01 · ReadThink
→
02 · Pick a toolAct
→
03 · Read the resultObserve
↻
04 · Until doneRepeat
that's the whole trick
10 · An agent
Why agents matter

Once an agent can code, it can do anything.

Code is the universal tool. Any task you can describe, an agent can now write the software to do. That's the unlock.

11 · The unlock
How agents reach into the world

CLIs. MCPs. APIs.

Three plugs. Three ways an agent does things beyond typing.

12 · The plugs
The real shift

The age of personal software.

For fifty years, software was made by a few, for the many. That just flipped. Describe a tool in plain English and have it. Not buy it. Not hire someone. Have it.

13 · Personal software
Laptops open →

Lab 04 · Build a browser extension

talkingtomachines.xyz/tools/extensions

copy the card · paste into AI Studio · export the zip · load unpacked

14 · Hands-on
The same loop, bigger jobs

What one agent carries.

Websites

This one. Every page, and every lab you used in Session 1.

Dashboards

Ask the business a question in plain language, get the chart.

Games

Rules, art, sound, playable by the evening.

Motion graphics

Title cards, explainers, animated numbers.

Video edits

Cuts, captions and reels from raw footage.

Deep research

A long report with sources, while you make tea.

15 · Full power
Where this is going

My agents will talk to your agents.

The next email you send might be drafted by your agent, read by their agent, replied to by theirs, and only surfaced to a human when there's a real decision. The humans stay in the loop. The loop gets a lot faster.

16 · Where this goes
Laptops open →

Lab 05 · Your three lines

talkingtomachines.xyz/tools/plan

four minutes · the wall fills as you go

17 · Hands-on
Workshop 1 · what you take away

You leave with.

  • The ladder, and an honest sense of where you sit on it.
  • Prompts, tokens, context and slop, learned by doing.
  • A de-slop system prompt in your own voice.
  • A skill you wrote yourself, plus where to find and share more.
  • A browser extension you built yourself.
  • A 90-day card on the wall: three lines, an hour a week, a buddy.
18 · Take away
Workshop 1 · end

Thank you.

The wall stays up. The hour is on your calendar.

Cultivated AI · Grow, don't Chase.
19 · Thank you