AI Is Making Me Smarter. Or Is It?

The Cognitive Debt Your AI Stack Is Accumulating

Something’s been gnawing at me. Partially because I can see how I could easily fall prey to the conflict I’m about to share.

I’ve been in this space long enough to have a front-row seat for multiple technology revolutions: the early web, eCommerce, mobile, social media and now AI. Each one reshaped how people work, communicate, and think. Each one had enormous upside. And each one came with a version of the same trap: the more powerful the tool, the easier it is to let it do the thinking for you.

With AI, that trap is particularly well-disguised. The output is so good. The answers arrive so fast. The experience of using it feels like being smarter, not less capable.

A few studies have come out recently that deserve an honest look.

The Research Is Starting to Measure What We Suspected

A 2025 study by researcher Michael Gerlich examined 666 participants and found a strong negative correlation between frequent AI use and critical thinking scores. The correlation coefficient was -0.68, which, if you’re not a stats person, is significant enough that it’s hard dismiss. The study is observational, so it doesn’t prove causation. People who rely more on AI may already think differently, or may be drawn to it for different reasons. But the association is real.

More pointed is a 2026 preprint from Shen and Tamkin (not yet peer-reviewed, but notable): software developers who fully delegated coding tasks to AI produced working code but scored 17% worse on conceptual quizzes afterward. They couldn’t debug what the AI had written. They had the output without the understanding.

Think about that. “Output without the understanding.”

And there’s a concept from cognitive research on what happens when AI consistently supplies full reasoning chains to users. Researchers call it the “hollowed mind” effect: familiarity with conclusions without the deep processing that builds actual expertise. You know the answer. You just don’t know why it’s the answer, or how to find it again without the tool.

It’s a literal dumbing down of ourselves.

This Isn’t New

The Google Effect, named by Sparrow and colleagues in 2011, showed that simply expecting access to search reduced spontaneous memory encoding, even before any search was performed. The availability of the tool changed how the brain bothered to process information. That was just search.

AI doesn’t retrieve information. It reasons. It synthesizes. It does the integrative cognitive work that is, as far as we understand it, how human expertise actually forms.

When a tool takes over the hardest parts of thinking, the parts your brain would otherwise have to struggle through, you don’t just save time. You also skip the struggle that would have made you better.

We can see this when some in the younger generation are unable to tell time from an analog clock. Why bother learning this basic skill when your phone spells it out for you?

The Two Modes

Here’s the framing I keep coming back to, both for myself and for how I think about sharing AI with other people.

There are two ways to use these tools.

1. Amplifier mode: You bring your judgment, your context, your hard-won pattern recognition. The AI accelerates the execution. You stay in the loop. You’re still doing the thinking; the tool is extending your reach.

2. Replacement mode: You hand the task over. The output arrives. You move on. You were present for the output and absent from the reasoning.

Replacement mode is seductive because it works, in the short term. The deliverable gets done. The code runs. The email goes out. Nobody sees the difference.

The cost comes later, and the effect does compound over time.

What This Means If You’re Building Something

If you’re an entrepreneur, executive, or anyone building something that depends on your ability to make good decisions, the question isn’t whether to use AI. Of course you should. I do. I use it to assist in my writing, project creation and image design.

The question is whether you’re using it in a way that keeps your reasoning muscles in use, or in a way that gradually outsources them.

Here’s the practical version:

- Use AI to draft, then rewrite at least one section yourself. Not to fix the AI’s work. To re-engage your own thinking with the material.

- When AI gives you a conclusion, ask it to show the reasoning, then pressure-test the reasoning yourself. The conclusion is easy to accept. The reasoning is where your judgment matters. Does it make sense to you?

- Don’t let AI handle your highest-leverage decisions. Use it for research, synthesis, speed. Reserve the judgment calls for yourself. Otherwise, you mind find yourself depending on some poor judgment.

- If you couldn’t defend the output without the AI in the room, that’s a huge red flag. Not a crisis, but information worth acting on.

The Honest Position

I’m not worried about AI in the abstract. I’m interested in the specific habits people are forming right now, in these early years, before those habits solidify.

For the adults reading this, the risk is mental atrophy. You built the reasoning capability; the question is whether you’re maintaining it.

For children who are growing up with these tools before they’ve built independent reasoning, the risk is different and starker. You can’t lose a capability you were never given the chance to develop.

Both are important. And both start with the same basic choice, made dozens of times a day: are you using the tool to think better, or instead of thinking?

What do YOU think? No AI replies, please!