All articles
AI Doesn’t Think. That’s Your Job.
July 7, 2026 
AI doesn’t think. Let’s start there.
It doesn’t understand the question you asked. It doesn’t know the answer it gave you. It has no idea whether what it produced is true, useful, or empty. It does one thing: it predicts which words should follow other words, at extraordinary speed, drawing on an ocean of human-generated text. That’s the whole trick.
And yet we talk about it as if it reasons. As if it knows. We ask it questions the way we’d ask a doctor or a trusted colleague, and we receive answers that arrive with the same smooth confidence whether they’re right or completely fabricated. That’s the gap that concerns me. Not the errors. The errors you can catch. The gap is what happens when we stop trying to catch them.
Think about what happens in your body when you stop using a muscle. It doesn’t disappear overnight. It atrophies slowly, quietly, until one day you reach for something and find you can’t hold it the way you used to. Judgment works the same way. Critical thinking is not a trait you’re born with and keep forever. It’s a capacity you build by using it, by struggling with hard questions, sitting with uncertainty, noticing when something feels off even before you can say why.
That friction is not a problem to be solved. It’s the work.
AI removes the friction. That’s what makes it useful. You ask, it answers, you move on. No struggle. No doubt. No sitting with the discomfort of not knowing. And because it sounds confident, fluent, and complete, the brain registers it as done.
But done is not the same as right. And fluent is not the same as true.
The danger isn’t that AI will deceive us deliberately. It has no intentions. The danger is that it sounds convincing whether it’s accurate or not, and the more we lean on that sound, the less we remember what truth feels like when we find it ourselves.
I’ve noticed this in how people use it. Not with alarm, but with curiosity. The question most people ask is “did AI give me the right answer?” The question they rarely ask is “what would I have thought if I hadn’t asked?” That second question is the one that keeps judgment sharp. It’s also the one we’re quietly learning to skip.
There is a word for what we are training out of ourselves: discernment.
The capacity to sense what’s off, what’s missing, what’s being smoothed over. It’s slower than prediction. It’s less convenient. It often produces doubt rather than certainty. And it’s one of the most distinctly human things we have.
We’re not at risk of AI becoming smarter than us. We’re at risk of us becoming less practiced at being smart.
That’s not a future scenario. It’s already visible in how we read, how we write, how we question, and increasingly, how we decide. The question worth sitting with is not what AI will do next. It’s what you do with the discomfort of thinking for yourself, even when a faster answer is one prompt away. Because that discomfort means something real is happening.
Over to you.
When did you last sit with a hard question long enough to find your own answer, before reaching for AI?
And what did that feel like?
Share in the comments.
Christian Kromme writes The Human Spark – Beyond AI, a newsletter about what it means to be human in a world being reshaped by technology.