Your request library

This lesson is still being tested: it improves with learners' feedback. If something is unclear, it is the lesson that needs fixing, not you.

By the end of this lesson, you will be able to

turn a successful request into a reusable template;
test a template on a new need, and improve it;
organize your requests following the method: SAY, DISCUSS, DO, CHECK, NOTE;
finish the module deliverable: 5 tested templates, and the story of an AI mistake that you caught.

Why it matters to you

A good request takes a few minutes to write. If you keep it, it saves you those minutes every time the need comes back. After a few weeks, you never start from a blank page again. This is the last step of the method: NOTE.

Before you start


You need	Your assistant, free account; your “My requests” note and your error logbook
Cost	€0
Duration	40 minutes
Reminder	No name, address or personal detail in a template: only [brackets]

Understand it in 3 minutes

The complete method, as you practiced it in this module:

Step	What you do	Lesson
SAY	The goal, the context, the form, and “It worked if…”	1
DISCUSS	Have it rephrase, ask for two options, ask for a critique, then give the green light	2
DO	Let it produce, then correct one point, ask for variants	3
CHECK	Open the sources, cross-check, reread with your criteria	3 and 4
NOTE	Keep what works, the decisions, and the next action	5 and 6

You don’t have to follow every step each time: a simple question only needs SAY. The more the work matters, the more of them you use.

A good template has four qualities:

it comes from a request that worked;
what changes each time is in [brackets], the rest is kept word for word;
if it produces something, it keeps its “It worked if…”;
it worked on two different needs.

Five good templates are worth more than fifty requests you never reread.

Do it together

1Step 1: Choose a request that worked

Open your “My requests” note. Take the request that got the best rating, for example the weekend one from lesson 1:

I want 3 weekend ideas, to decide tonight with my friends.
What you need to know: we are 4 adults, traveling by car from Lille, with a budget of €150 per person, and one person in the group can't walk for long.
Form: 3 proposals of 5 lines each, with the travel time.
It worked if the 3 ideas fit the budget and are less than 2 hours away by road.

2Step 2: Turn it into a template, with the AI’s help

Here is a request that worked well. Turn it into a reusable template: put in [brackets] what changes from one time to the next, and keep the rest word for word. Then tell me in one line when to use it.

[paste your request here]

You get something like:

I want [number] ideas for [what], to [decide what, or for whom].
What you need to know: [who, how many, budget, constraints].
Form: [number] proposals of [length], with [the key information].
It worked if [criterion 1] and [criterion 2].

Reread: did the AI keep the criteria? Did it leave a personal detail? Fix it if needed.

3Step 3: Test it on another need

Fill in the brackets for a different need, for example gift ideas for a colleague who is leaving, with your budget and your constraints. Send it, then rate the result out of 5. If it doesn’t reach 4, improve one line of the template and try again.

4Step 4: Organize

Open your request library (resource below) and put your template in the “My 5 tested templates” section, with its two tries and their ratings.

Your turn

The module 2 deliverable, to finish in your library (it takes more than the lesson’s 40 minutes: spread it over a few days if needed):

5 templates tested on your real needs, each on two different needs. At least one template for SAY, one for DISCUSS, one for CHECK.
The story of an AI mistake that you caught, based on your error logbook (lesson 4): what you asked, what it answered, how you spotted it, what was true and where you read it, and what you do now.

If the AI did not make a mistake during your checks, redo one of the module 1 tests (lesson 2) that can bring out a mistake, for example the word count or the request for books on a subject you know well.

Failing that, tell the story of a check that confirmed the answer, and what allowed you to be sure: the quality of the check is what matters.

Resources to download

My request library: your settings, ready-to-copy requests, your 5 templates and the story of your mistake.

Check your work

It worked if:

your library contains 5 templates, each tested on two different needs, with a rating of at least 4 out of 5 on both tries (after improvement if needed);
each template that produces something keeps its “It worked if…”;
no template contains a name, an address or a personal detail;
your mistake story says precisely how you caught it.

Common mistakes:

keeping requests without ever retesting them: a request that worked once is not yet a template;
templates that are too specific (no brackets) or too vague (no context or criteria);
a library you can’t find: store it where you can find it in 10 seconds;
a mistake story without the “how I spotted it”: it is the part that will help you.

Keep it and pick it up later

Your library is your main working tool for modules 3 and beyond.
When a request works well, add it right away; once a month, remove what you no longer use.

Going further

optional

Your library works with any assistant: if you change tools or plans, you keep your requests. The Choosing your tool page compares the Claude and ChatGPT plans, with dates.
In module 7, you will see how an agent can keep a reusable way of doing things as a skill, a bit like a template from your library.

New words

4 words explained

Word	What it means
Request template	A request that worked, where what changes is in [brackets]
Brackets	The blanks to fill in a template: [what], [for whom], [budget]
Request library	The place where you keep your tested templates, your settings and your caught mistakes
Deliverable	The concrete result you keep at the end of a module