Relearning Thinking After Asking ChatGPT First Too Often

5 prompts to activate your own brain before you activate the machine

Recently, I read an article about someone stopping ChatGPT for 30 days. The article was dramatic, maybe a little too dramatic, but the core idea was hard to ignore: many of us are losing our tolerance for mental friction.

I Stopped Using ChatGPT for 30 Days. What Happened to My Brain Was Terrifying.

91% of you will abandon 2026 resolutions by January 10th. Here’s how to be in the 9% who actually win.

Level Up CodingKusireddy

The comments were even more interesting.

Some readers agreed completely. Others pushed back: “Should we stop using cars too? Phones? Google? The internet?” - Fair point.

The goal cannot be to reject useful tools. That would be silly. ChatGPT is useful. Sometimes annoyingly useful. Like a colleague who is both brilliant and slightly too available. And dangerously useful so that we forget to think ourselves.

The real problem is not that we use AI, but that we often ask AI before we have asked ourselves.

We have learned prompt engineering for machines, but we are forgetting prompt engineering for our own brain.

We forget to prompt ourselves

Look at how carefully we talk to AI:

Act as a senior strategist.
Challenge my assumptions.
Think step by step.
Give me ten alternatives.
Use first principles.
Find the hidden flaw.
Explain it simply.

Then we turn to our own brain and talk to it like this - if at all:

Think.
Faster.
Why nothing?
Bad brain.
Open ChatGPT!!!

No wonder the machine gives us the better answers: We gave it better questions.

This is the small tragedy of the AI age: we did not only outsource answers. We outsourced the habit of asking ourselves good questions.

But this is also good news: the fix is not all too complicated:

Before prompting AI, prompt yourself.

Five to ten minutes and a notebook or a blank text file might be enough.

The rule is simple:

Brain first. AI second. Brain last.

First, your own rough thinking.
Then AI for critique, expansion, and comparison.
Then your own synthesis, so the insight becomes yours.

This article gives you five brain prompts for exactly that.

Prompts that make your own mind wake up before the machine takes over the room.

The workflow: Human → AI → Human

Here is the whole system for you to not forget:

1. Prompt your own brain.
2. Write the rough human answer.
3. Ask AI for a second perspective.
4. Compare your answer with AI’s answer.
5. Synthesize the best version.
6. Explain/Summarize it without AI.

Let’s unpack what is happening here.

Step one is important because your brain needs retrieval. If you never ask it to search, connect, remember, and struggle, it gets lazy. Not broken. Lazy. Like a dog that realizes the human will always pick up the ball.

Step two creates a visible artifact. A messy outline. A bad paragraph. A list of possible ideas. Now AI has something to respond to. (You might prompt AI without it first - to see how it would respond without any of your input!)

Step three lets AI do what AI is excellent at: expanding, criticizing, reframing, generating alternatives.

Step four is where learning happens.
You compare.
You notice what AI saw and what you saw.
You notice where AI was stronger.
You also notice where your own weird little idea was more alive.

Step five turns the pile into insight.

Step six is the test:

If you cannot explain the final result without AI, you did not understand it yet. You only rented fluency.

Stolen ideas are amazing, but they are also dangerous.

They let you sound smarter than your actual understanding.

The goal is not to sound smart. The goal is to become harder to fool, harder to confuse, and harder to replace.

So let’s prompt the brain properly!

Prompt 1: “What do I already know before I ask?”

This is the anti-reflex prompt. Use it tos top your hand move toward ChatGPT before your brain has even taken its shoes off.

Ask yourself:

What do I already know about this?
What do I suspect is true?
What examples do I remember?
What would I try first?
Where exactly am I stuck?

Or simply:

What are my raw/first thoughts about/to this?

Do this for five minutes.

The purpose is not perfection. The purpose is activation of your brain.

Your brain is associative. It often needs a first hook. Once you write one idea, another appears. Then a memory. Then a doubt. Then a connection. Then the beginning of a real thought.

This prompt is especially useful for writing, coding, strategy, studying, and problem solving.

For example, instead of asking:

ChatGPT, write me an article about AI and deep work.

Start with:

What do I already think? What are my useless (or useful) first thoughts about this question? What is obvious and what might be not so obvious?

Your brain needs to enter the conversation, before you ask AI first!

• We ask AI too early.
• The problem is not AI, but skipping the first struggle.
• We prompt AI better than ourselves.
• Need a method: self-prompt, AI prompt, comparison, synthesis.
• Final test: can I explain it alone?

AI companion prompt

After you write your own rough answer, use this:

Here is my rough thinking:

[paste notes]

Do not rewrite it yet.

First, tell me:
1. What is strongest here?
2. What is unclear?
3. What am I missing?
4. What would a skeptical reader object to?
5. Which idea is most original and worth developing?

And of course - actually more interesting would be to ask it yourself - not AI.

Prompt 2: “What am I assuming that might be false?”

Most bad thinking is not caused by stupidity, but by invisible assumptions wearing a fake mustache.

We assume the article must be long.
We assume the code must be complex.
We assume the business idea needs a full platform.

This prompt is a quiet little crowbar.

Ask:

What am I assuming?
Which assumption might be false?
What would change if the opposite were true?
What constraint did I invent myself?
What would this look like if it were simpler?

For the ChatGPT problem, we might assume:

Assumption: The solution is to quit AI.
Maybe false: The solution may be to use AI later, not less.
Assumption: AI makes people lazy.
Maybe false: Bad sequencing makes people lazy.
Assumption: Human thinking and AI thinking compete.
Maybe false: They compound if used in the right order.

Now the article becomes better.

Instead of another “AI is destroying us” sermon, we have a more interesting thesis:

The problem is not using ChatGPT. The problem is letting ChatGPT be the first mind that touches the problem.

That is sharper and came from questioning assumptions.

AI companion prompt

Use this, if you want to involve ai in this step. Otherwise try to answer yourself with your brain!

Here is my current idea:

[paste idea]

Find the hidden assumptions.

For each assumption:
1. State it clearly.
2. Explain why it might be wrong.
3. Give the opposite assumption.
4. Show what new idea appears if the opposite is true.

Perhaps this is one of the best uses of AI: Assumption surgery.

Prompt 3: “Give me ten bad ideas.”

Your brain often blocks ideas because it wants them to be good immediately.

Creativity usually needs quantity before quality.

The first ideas are obvious.
The second ideas are borrowed.
The third ideas are trying to impress someone on LinkedIn.
The tenth idea may finally be interesting.

So ask:

What are ten bad ideas?
What are ten strange ideas?
What are ten opposite ideas?
What are ten ideas I would normally reject too quickly?

Why bad ideas? Because bad ideas lower the gate.

Your inner critic relaxes. The guard dog stops barking. The weird useful thought sneaks in through the side door.

For this article, ten bad ideas might look like this:

1. Never use AI again.
2. Only use AI on Sundays.
3. Ask AI only after coffee.
4. Make ChatGPT insult you if you ask too early.
5. Put a 10-minute delay before opening AI.
6. Use a notebook as a password to AI.
7. Force yourself to write three guesses first.
8. AI can only critique, not create.
9. Every prompt must start with “Here is my own attempt.”
10. Use AI like a gym coach, not like a replacement body.

Some are silly, but number 9 is useful.

Every prompt must start with: "Here is my own attempt."

That is a real rule.

Number 10 is also useful: Use AI like a gym coach, not like a replacement body.

That metaphor may survive into the article. The bad ideas did their job: They found the good ones.

AI companion prompt

After your own list, ask AI (or your own brain):

Here are my ten bad ideas:

[paste list]

Find the hidden good ideas inside them.

For each one:
1. What useful principle is hiding here?
2. How could it become practical?
3. Which idea is more original than it first looks?
4. Which one could become a memorable line or framework?

AI is excellent at this. You bring the strange raw material. AI helps refine it. That is collaboration.

Prompt 4: “What did AI see that I missed — and what did I see that AI missed?”

This is the comparison prompt.

It is where AI becomes a teacher instead of a vending machine.

Most people ask AI for an answer and stop.

That is like hiring a tutor, letting them solve the problem, and then saying, “Wonderful, I now possess mathematics.”

No.

You possess a solved page. Learning begins when you compare. First write your answer. Then ask AI for its answer. Then compare both.

Use this structure:

My answer:
AI’s answer:

What is better in mine?
What is better in AI’s?
What did AI miss?
What did I miss?
What feels generic?
What feels alive?
What should be combined?

This is powerful because AI will often be better than you in certain ways.

It may be more structured, more complete, more balanced.
Faster at generating examples, or better at seeing obvious objections.

Good. Steal the useful parts!

But your own ideas may be better in other ways.

More personal, more specific, more unusual, more weird,
more connected to your actual experience, and especially: emotionally honest.

Protect those!

AI often gives you the plausible center or sometimes the too obvious, generic answers.

Your brain often finds the strange edge. Many great ideas live at the edge.

For this article, AI might suggest: “Use AI as a tool, not a crutch.”

True. Useful. Also familiar.

Your own brain might say: “We prompt AI better than we prompt ourselves.” (And indeed it was what my brain told me.)

That is fresher. That is the one to keep.

AI companion prompt

Use this (but also ask your brain!):

Compare my answer with your answer.

Create a table with:
- Idea
- Source: me or AI
- Strength
- Weakness
- Originality
- Practical value
- Keep, cut, or combine

Then answer:
1. What did I notice that you missed?
2. What did you notice that I missed?
3. What is the best synthesis?
4. Which idea should become the central thesis?

This prompt is gold, because it forces you to judge.

Judgment is thinking! And we need to practice it.

Prompt 5: “How do I lock this insight into memory?”

Solving a problem feels good. Keeping the insight is harder.

We often have a breakthrough, feel clever for 14 minutes, then forget the whole thing by dinner because a notification, a sandwich, and three unrelated tabs attacked our working memory.

So after you solve something, ask:

What did I learn?
What is the principle?
Where else can I use it?
What mistake should I avoid next time?
How would I explain this simply?
What question should I ask myself tomorrow?

The brain remembers better when it retrieves, connects, explains, and reuses information. You do not keep insight by admiring it once. You keep it by handling it again.

A useful format:

Problem:
What I first thought:
What AI added:
What I finally understood:
Where I can use this again:
One sentence I want to remember:
One follow-up question:

For example:

Problem:
How to use ChatGPT without becoming dependent.

What I first thought:
Use AI less.

What AI added:
Use AI after self-thinking and before synthesis.

What I finally understood:
The sequence matters more than the tool.

Where I can use this again:
Writing, coding, studying, decision making.

One sentence I want to remember:
Prompt yourself first, prompt AI second, prove you understood last.

One follow-up question:
Where else am I asking tools before asking myself?

That is how an idea becomes reusable.

AI companion prompt

Use this after any serious AI session (or ask your brain again!):

Help me turn this conversation into learning.

Extract:
1. The main insight.
2. The general principle.
3. Three situations where I can reuse it.
4. One mistake I made in my initial thinking.
5. One question I should ask myself next time.
6. A short memory phrase.
7. A 24-hour recall question to test myself tomorrow.

Actually in drawing insights, your brain is better than ai, because it knows what you knew before or didn't knew before! And it can generate the future situations better, in which you might need what you learned.

This question also anticipates (prepares) your brain for future encounters!

The five brain prompts in one place

Here is the compact version.

Before AI:

1. What do I already know before I ask?
2. What am I assuming that might be false?
3. What are ten bad ideas?
4. What did AI see that I missed — and what did I see that AI missed?
5. How do I lock this insight into memory?

Number 4 comes after AI, and number 5 comes after the whole session.

So the whole workflow is:

Before AI:
1. What do I already know?
2. What am I assuming?
3. What are ten bad ideas?

With AI:
4. What did each of us miss?

After AI:
5. How do I lock this insight in?

A small rule that changes everything

From now on, try this:

Never open ChatGPT without prompting your own brain.

Open it with a rough attempt, a bad attempt, a confused attempt, but an attempt.

Start your serious prompts like this:

Here is my own thinking first:
...

You can add your first own thinking or leave it out from the prompt (then just write it somewhere else!).

This one habit protects your brain.

It forces retrieval before assistance.
(It gives AI something concrete to improve, if you involve it to the AI prompt.)
It makes comparison possible.
It helps you notice what you actually understood before asking AI.
It turns every AI session into a practice, not just consumption.

This is the difference between using AI and being used by the convenience of AI.

The final test: can you explain it alone?

After ChatGPT helps you, close it.

Then explain the idea without its help out of your own memory.
Or at least explain/repeat what you want to keep in memory for the future.

Ask yourself:

Can I explain the problem/the solution/ how/why it works?
Can I give an example/objection?
Can I apply it somewhere else?

If yes, you learned. If no, you only finished.

Finishing is sometimes enough. We are adults. We have invoices. And emails have to be sent.

But for developing skills, finishing is not enough.

You want to become more capable after using AI than before using it.

The standard shouldn't be: “Did AI give me an answer?”

But: “Did this interaction improved my brain/skills?”

Use ChatGPT! Just do not disappear into it.

I am not quitting ChatGPT. That would be absurd.

I use it for writing, coding, brainstorming, editing, learning, debugging, and occasionally asking whether a sentence sounds human or like it was assembled in a conference room by three SaaS founders and a scented candle.

AI is useful. But the better I get at using it, the more I think the first step should often be:

• use first your brain AI-free.

And after it:

• learn from/with AI!