Path · From zero to your own AI agents

Learn AI. In the order I learned it in.

It took me two years to figure out how all of this fits together — for nothing. If I had to do it over, I'd take this route. Four steps that build on each other, 13 articles already written, all tested on my own business before they landed here. Take them in order, skip what you already know.

4steps 13articles ~2h30of reading €0to read it all
Before you dive in

Your real questions.
The straight answers.

Everyone asks the same ones. Let's settle them now, so you start with a clear head.

Do I need to know how to code?

No, and that's the whole point. For 90% of what you'll do, zero lines of code: you write in plain language, the AI does the work. Claude Code, the most "technical" tool, shows up in step 02 — and even there I hold your hand from start to finish.

Isn't it too late?

No, you're arriving at exactly the right moment. Half of everyone started this year. You're not behind — you're just showing up when the tools have finally become simple. The ones who started two years ago, like me, struggled for nothing.

Will AI replace my job?

What replaces you is someone who knows how to use it, not the machine. Learning to use it is exactly how you stay in the game. That's the goal of this path: putting you on the right side of it.

How long before I'm comfortable?

A few hours for the basics, a weekend to be self-sufficient. Not six months. You can even do your first genuinely useful thing today — the block right below is there for that.

I won't get any of it, it's too technical?

That's exactly what this path avoids. Zero jargon, every word explained, a glossary if a term trips you up. It's arranged in the order you actually learn it. If a sentence still loses you, write to me: I'll rewrite it.

0 / 4 steps completed
Check off your steps as you go — it's saved on this device.
📬 The path, guided version

Would you rather be walked through it?

Get the path over 7 days: one short email a day, in the right order, so you move forward without second-guessing. Free, and you unsubscribe in 1 click.

7 days, 1 short email a day. Zero spam, unsubscribe in 1 click.
01
Step 01 · The foundations

Lay
the basics.

If you've never really used AI beyond ChatGPT, start here. Three reference points to understand what we're working with, get the 8 core words down, then keep the A-Z glossary on hand whenever a term trips you up.

3landmarks ~35 mintotal LevelBeginner PrerequisitesNone
⚡ Do it now · 2 min

Your first real prompt, right now.

Before you even read anything: open ChatGPT, Claude or Le Chat (all free), paste this, and replace the part in orange with whatever you want to understand.

The prompt
You are a patient teacher who explains things without jargon.
Explain to me simply [what you want to understand — e.g. GDPR, an electricity bill, what AI is].
Give me: 1) a one-sentence definition, 2) an example from everyday life, 3) the one thing to remember.
What you'll see

An answer in 3 clear points, no fluff. And most of all: you just wrote a real prompt — a role ("patient teacher"), a precise request, an imposed format. That's the whole craft. The rest of the path is just refining that.

— Next step · 02
Move to Claude Code
02
Step 02 · Move to Claude Code

Hand
it the keys.

Claude Code is Claude in your terminal. It reads your files, it writes code, it runs commands, it fixes its own mistakes. This is where AI goes from a chat tool to a coworker who works for you. Install, loops for tasks that run in the background, 6 official plugins, and GitHub as your safety net.

4reads ~41 mintotal LevelBeginner ToolsMac Terminal
— Next step · 03
Build an agent
03
Step 03 · Build your agents

Put it
to work for you.

An agent is Claude with a clear mission, some tools, and the right to act on its own. This is where AI goes from "I'm thinking about it" to "it runs by itself". We start with the minimum theory, then two concrete tutorials you can replicate tonight: an agent that sorts your emails, one that goes through your contracts. Then a notch above, a pro-level local agent.

4reads ~48 mintotal LevelIntermediate Goal1 agent running
— Next step · 04
Go further
04
Step 04 · Go further

Dig
in deeper.

The basics are down, your first agents are running. From here, two directions: real cases (what I've built, with numbers and lessons learned) and frontier research (Karpathy's work in plain terms, to understand where the field is heading). Take whatever interests you.

5reads ~55 mintotal LevelCurious Readin whatever order speaks to you
Real case · 04.1 8 min

A sales tool,
built over
3 evenings.

— Netflix in the background · Not a line written
What it's about

Shirley (an insurance account manager) asked me for a hand. Three evenings, a sales tool now used in client meetings. The repeatable process, not the magic.

Real case · 04.2 9 min

My watch
for tomorrow morning,
prepped overnight.

— 100+ sources · Auto brief · Subagents
What it's about

The real pipeline behind my AI Playbook and Business Radar newsletters. What runs overnight, what gets filtered, what's left for me in the morning. No hidden costs.

Research · 04.3 15 min

Karpathy,
in plain terms.

— 14h40 of video summed up in 15 min
What it's for

Andrej Karpathy put together a series of 8 videos that goes from "what's a neuron" all the way to building a full GPT-2. I give you the 15-min version — you keep the essentials without spending 14h on it.

Research · 04.4 12 min

Karpathy's
self-improving
agents.

— An agent that edits its own code
What it's about

A repo that kicks off a 5-min training run, lets an agent edit the code, compares the result, keeps it or reverts. Indefinitely. "Research on a shoestring", within reach in 30 min at home.

Research · 04.5 12 min

The LLM Wiki,
your second
brain.

— An idea · Not a tool
What it's about

Karpathy's idea: a personal wiki read by an LLM to answer in your place. When it's worth it, when it's overkill (Projects in Claude.ai is often enough), how to go about it.

Tutorial · 04.6 15 min

Open source,
explained for non-devs.

— What it is · Why · Which tools
What you'll understand

The word is everywhere since AI took off. What it is, how companies make a living giving things away for free, why Mistral and DeepSeek are changing the game. 30 tools to try tomorrow. No jargon.

Tutorial · 04.7 20 min

Run a real AI
on your own
machine.

— No subscription · No cloud
What you'll do

Install an AI on your computer in under an hour, with Ollama and Open WebUI. Chat with your own documents (RAG). 6 use cases by profession. Step-by-step tutorial for Mac/Windows/Linux.

End of the path · Stay in touch

You've read it all.
Now what?

My AI Playbook newsletter goes out every Friday morning. My subagents scan 100+ sources during the week, I get the brief, you get the exact same email I do. That's how I stay up to date, and that's how you can stay up to date too. Unsubscribe in 1 click if it's not useful to you.

See the newsletter