How to Use AI Without Losing Your Ability to Think

The scariest thing AI can do to you is not take your job. It is to slowly, quietly take your ability to think — and to feel helpful the entire time it does it.

I am not an AI doomer. I use these tools every day and they make me faster. But I have watched something happen in myself and in people I respect: a kind of mental softening. The reflex to reach for a prompt before reaching for a thought. The vague unease of realizing you can no longer hold an argument in your head the way you used to. That is the risk worth talking about, and almost nobody is.

The use-it-or-lose-it problem

Your brain is not a hard drive. It is more like a muscle, and it follows the same brutal rule: what you do not use, you lose.

Researchers call the mechanism "cognitive offloading" — the act of handing a mental task to an external tool. This is not new. We offloaded memory to writing, arithmetic to calculators, navigation to GPS. Offloading is often smart. The problem is what happens when you offload the core work instead of the busywork.

When you let AI do the thing that actually builds the skill — the recalling, the reasoning, the weighing of evidence, the wrestling a messy idea into a clear sentence — you do not get the skill. You get the output without the adaptation. And the next time, the task feels a little harder to do alone, so you offload again. That loop is how a capable mind gets quietly hollowed out while producing more work than ever.

Most people will not notice it happening. The deliverables look fine. The decline is invisible until the moment you need to think without the tool and find the muscle has gone slack.

Amplify your thinking, or avoid it

Here is the whole article in one distinction.

There are two ways to use the exact same tool, and they lead to opposite destinations.

Avoiding thinking: "Write me an argument for why we should expand into Kenya." You read it, it sounds reasonable, you ship it. You now hold a position you never actually reasoned your way into. You cannot defend it under pressure because you never built it.

Amplifying thinking: "Here is my argument for expanding into Kenya. Here are my three reasons. What is the strongest case against each one? What am I not seeing?" You did the thinking. The AI stress-tested it. You walk away with a better version of your own judgment — and you still own it.

The first use makes you dependent. The second makes you formidable. The tool does not decide which one you get. You do, every single time you open the prompt.

The trap is that avoiding thinking feels exactly like being productive. The output is similar. The cost shows up later, somewhere you are not looking.

Practical rules I actually follow

Principles are useless without practice. These are the rules I use to keep myself on the right side of that line. None of them require willpower heroics. They are mostly about sequence — doing the human work before the machine work.

1. Struggle with the problem first

Before I ask AI anything, I sit with the problem long enough to form my own rough answer — even a bad one. The struggle is not wasted time before the "real" work. The struggle is the work that builds the mental structure. AI is allowed to enter after I have something to react to, never before.

2. Use AI to challenge you, not to replace you

My most valuable prompts are not "do this for me." They are "argue with me." Ask it to find the holes in your plan, the counterevidence to your belief, the assumption you are treating as fact. Used this way, AI becomes the sharpest sparring partner you have ever had — and sparring builds the muscle instead of replacing it.

3. Write your own first draft

Whatever it is — the strategy, the essay, the difficult message — I write the first version myself, ugly and incomplete. Then I bring AI in to critique and refine. The first draft is where thinking happens. If you let AI write it, you have handed away the one part that was actually yours.

4. Verify before you trust

AI states wrong things with the same fluent confidence it states right ones. It does not know when it is guessing. So I treat every factual claim, statistic, citation, and "this is how it works" as unverified until I check it against a real source. Outsourcing your thinking is bad. Outsourcing your skepticism on top of it is how you confidently repeat things that are false.

5. Keep doing hard mental work on purpose

This is the one nobody wants to hear. You have to deliberately keep doing some hard cognitive work that you could have offloaded — for the same reason you lift weights you do not strictly need to lift. Do mental math sometimes. Read the long, difficult thing without summarizing it. Hold an argument in your head and turn it over. The point is not efficiency. The point is keeping the machine that makes you you in working order.

Why thinking becomes more valuable, not less

Here is the part that should make you optimistic, and a little ambitious.

As AI gets better, answers become cheap. Anyone can generate a competent essay, a plausible plan, a passable analysis, in seconds. When something becomes abundant, its market value collapses. Competent output is on its way to being worth almost nothing, because everyone has infinite access to it.

What does not become cheap is judgment. Knowing which question to ask. Knowing when the confident answer is subtly wrong. Knowing what matters and what is noise. Taste, discernment, the ability to take responsibility for a decision — these become more scarce and more valuable precisely because raw answers are now free.

So the person who used AI to avoid thinking has trained themselves out of the only skill that is appreciating. And the person who used AI to amplify their thinking has spent the same period getting sharper. Same tool. Two completely different futures.

The whole game is making sure you are the second person.

Frequently asked questions

Isn't using AI to write or solve things just being efficient?

It depends entirely on what you offload. Handing over busywork — formatting, boilerplate, first-pass research you will verify — is genuine efficiency. Handing over the core reasoning, the judgment, the first draft of your own thinking is not efficiency, it is atrophy wearing a productivity costume. The output looks the same now. The difference shows up in six months.

How do I know if I'm becoming too dependent on AI?

Try doing a meaningful thinking task with no AI at all and notice how it feels. If writing without it, deciding without it, or reasoning without it now produces anxiety or feels markedly harder than it used to, you have drifted into dependence. That spike of discomfort is the signal to deliberately do more unassisted work, not less.

Should I just avoid AI to protect my thinking?

No. That is overcorrecting, and it leaves real capability on the table. AI is an extraordinary sparring partner and amplifier when you do the thinking first and use it to challenge and refine your work. The goal is not abstinence. It is sequence: human work first, machine work second, and never outsourcing the part that builds the skill.

What's the single most important habit to keep?

Write your own first draft. Whether it is an essay, a strategy, or a hard decision, produce the messy first version yourself before AI touches it. The first draft is where the actual thinking happens. Keep that one human and you preserve most of what matters, even if you use AI heavily for everything after it.

Why would thinking matter more as AI gets smarter?

Because AI makes answers abundant, and abundant things lose value. What stays scarce is judgment — knowing which question to ask, spotting the confident answer that is wrong, deciding what actually matters, and owning the decision. As raw output approaches free, the human ability to think well becomes the rare and valuable thing.

The bottom line

AI is not going to think for you against your will. It is going to think for you whenever you let it — and it will feel helpful every time, which is exactly what makes it dangerous. The decline is never dramatic. It is a thousand small surrenders that each felt reasonable.

You do not have to make those surrenders. Struggle first. Use the tool to argue with you, not for you. Write your own draft. Verify what it tells you. And keep doing hard mental work on purpose, even when you do not have to. Do that, and AI becomes the best thinking partner you have ever had instead of the thing that quietly replaced the thinker.

Use it to get sharper. Most people will use it to get softer. That gap is the whole opportunity.

Further reading on this site

• How to Learn Anything Faster With AI
• Deep Work in the Age of Infinite Distraction
• Browse Personal Growth

If this resonated, subscribe to the newsletter for more honest, practical writing on thinking well in the age of AI.