Dear , let me introduce you to my mom:
She’s a retired pharmacist.
So she has no reason to care about AI.
But being my (Jewish) mom, she wants to understand what her son is teaching to 900,000 readers. And I want to finally answer the question, “Who’s Claude?”.
(Claude is a French name, I’m French; it checks out)
For my mom, AI is not Claude. It’s not even ChatGPT. AI is just Gemini (because she has an Android and they are pretty aggressive with Gemini).
Claude is this small to the rest of the world:
So for this newsletter, I will teach you Claude the same way I taught my mom Claude. And by the end, you will know what matters, your wrong assumptions about it, and even use it the way so-called Claude pros do.
But before we start, I’d love you to send this article to your mom.
It’s free, and it will bring her in the AI world:
This guide works with every AI, ChatGPT, Gemini, or Grok. Soon, they will be so similar it will be like picking up sneakers.
1. So that’s Google, right?
No Mom, Claude (and AI) is not Google.
It’s tempting to think it is because you search for something and it magically answers. But it’s closer to your phone trying to predict the next word.
This is how AI works.
Let’s take the sentence “This is how AI finds the next […]”
Claude goes through 4 steps to determine […]:
My mom then asked, “But how does it know which word?”, so I had to explain very simply how AI gets trained. It read so many books and internet discussions that it got a good feeling for how sentences are made.
Was it done legally and ethically? Meh.
Anthropic trained on millions of single-unit books, and then burned them:
But so did ChatGPT when they trained their AI. And now new AIs are distilling other AIs (it means they are training themselves on other AI answers).
If we must learn something from history, it’s that technology can’t be stopped. It can be contained, regulated, safely deployed - but never stopped.
So I am happy my mom retired.
Because she never enjoyed tech so much. And the world is dividing in two: either you’re extremely good at handling frontier technology (like AI), or you’re so likable people keep coming back to your shop (that’s my mom at the pharmacy).
But the “likable = job” isn’t so obvious.
For example, we used to have porter as a job:
I’m sure they were very nice and likable, like my mom - but then we invented wheels on luggage, and the roads became easier to navigate.
Technology replaced likable people.
If your mom needs to get better on the technology side, with AI, let’s continue the newsletter.
2. Her first prompt.
Prompting is easy, Mom, you just need to structure it this way:
I need [goal] that I will consider achieved once [success criteria].
I am [context about me, my situation].
If necessary, ask me clarifying questions first.
I promise, you won’t need more than this.
Let’s take an example: my mom had to pay for a notary to plan her succession. It was a technical and legal question (that is costly to ask!!).
It involves her property, her marital status, the fact that she has kids (we’re four), and tons of new French laws no one reads and stays on top of.
So we started with this prompt:
I need a plan to pass my apartment to my four children with minimal tax, which I’ll consider achieved once I have the best options under French law.
I am 61 (she looks like she’s 30 still), retired, in Paris, 4 children, one apartment. I want to keep living in it after transferring it.
If necessary, ask me clarifying questions first.
Now the truth is, my mom can’t type this.
She’s a very slow typer.
So I showed her Wispr Flow, a dictation tool. We simply talked like she would to her notary.
At some point she got scared: “What if I don’t say enough?”. But that’s Claude’s job. That’s why we added, “If necessary, ask me clarifying questions first.”
I love Claude for this reason. It creates an easy form:
10 minutes in, and my mom got every answer the notary told her… without the cost! She couldn’t believe it.
“It answers so well! Maybe we should share it with the notary?”
I’m not done, Mom.
Searching and finding answers is one step.
Let’s create the right strategy & write the contract:
3. More than answers.
From that point onward, my mom understood what makes AI different than Google.
You can literally write the contract:
A complete strategy document.
The pre-filled contract a notary can review.
All of this, almost for free, within <10 minutes.
I told my mom, “Imagine if you started the conversation with the notary having these documents first. Would it be the same price?”
I believe not.
That’s how you expand someone’s brain about AI: they finally feel how it widens their options. They are now capable.
But my mom needs more. She’s a tough cookie.
“How can we be sure of this ‘Claude’? What is its source?”
You’re right. It’s called dealing with hallucinations.
4. Dealing with hallucinations.
‘Hallucination’ is what we call AI making things up.
We say that “The AI hallucinated [something].”
And it’s annoying.
In a recent paper, they found that if you show the AI a fancy-looking chart, it thinks it’s real. The AI doesn't check whether the numbers are true. It just trusts that the picture looks official.
So naturally, my mom got scared.
But you can force Claude or other AI to cite their sources and pick official ones only. Just add:
Answer only from [DOMAINS]. Cite the exact reference and link
for every figure. If it’s not there, say “not found”. Don’t fill from memory.
It looks like this for my mom:
Answer only from legifrance.gouv.fr and bofip.impots.gouv.fr.
Cite the exact article and link for every figure. If it’s not there, say “not found”. Don’t fill from memory.
And it looks like this inside Claude:
And this is true with every “but AI is bad” comments → people are one-sided because they had one bad experience or they heard one bad story.
Yes, hallucinations exist.
Yes, you can fix it.
Yes, if you use a public Claude and you share private data, it has some risks.
Yes, if you use a Business Claude, they are SOC 2 compliant, so it can’t leak that data. The same way Microsoft or Google works.
Once she felt confident, she asked me the final question:
5. Should I open Claude?
No, you don’t, Mom. You’re retired.
Enjoy your grandchildren.
That being said, you will most likely use a product from Claude without even noticing. Because it’s so easy to build anything with Claude.
Let’s build one together: remember the strategy and the contract we wrote for her specific situation? What if we turned it into a website so any mom in her situation could have the results?
“We can do this?” Yes, we can:
6. How I use Claude.
Less and less like I taught my mom.
More and more with Claude Code instead.
Now I’m not technical. I don’t know how to code. But I’ve been spinning up live websites almost every day for any use case with this simple prompt: