The AI Paradox
The smarter our tools get, the more it matters that we keep thinking.
🧠 The oldest worry in a new form
Socrates worried that writing would weaken memory. Later, people made similar claims about calculators and spell-check. Now that AI can draft our emails, write code and summarise books, the question is back: is AI making us less capable?
That is the paradox. AI can make people more capable, and it can also tempt them to stop practising the skills that make them capable. The research on this is still young and points in both directions, so the honest answer is it depends on how you use it.
⚖️ Two fair readings
Thoughtful people disagree here. Both sides have a real point.
The optimistic case
- Faster drafts, summaries and first passes free up time for harder work.
- Anyone with an internet connection can get patient, personalised explanations.
- Small teams and solo creators can build things that once needed a whole department.
The cautious case
- Skills you stop practising can fade, as many people found with navigation apps.
- Fluent, confident output can hide mistakes, so errors are easy to miss.
- Jobs, trust and the quality of online content may change faster than society adapts.
🎛️ The deciding factor: how you use it
The same tool can sharpen you or dull you. The difference is agency. Tap a tab to compare.
The copy-paste router
"Write it for me. I'll paste it and publish."
- Skill drift: if you never build an argument yourself, you get less practised at it.
- Borrowed confidence: a polished answer can feel like your own understanding when it isn't.
- Unchecked output: accepting answers without questioning them lets errors through.
The editor-in-chief
"Give me three outlines. I'll pick one, rewrite it in my voice and check the facts."
- A sparring partner: you ask AI to challenge your ideas, then decide what holds up.
- A tutor: you ask it to explain what you don't know, then test yourself.
- Judgment in the loop: reviewing and correcting its work is itself thinking practice.
🐎 The "centaur" lesson, with a catch
In the 2000s, "freestyle" chess teams that paired a human with a computer often beat both lone grandmasters and lone engines. The human chose plans and caught the machine's blind spots, while the machine checked the tactics.
The catch: chess engines have since become so strong that the human edge there has mostly disappeared. So the centaur is not a guarantee. It is a reminder that the pairing only works while the human is genuinely contributing.
🎮 Quick test: which prompt would you send?
You need to write a newsletter about a new industry trend. Pick your approach.
📊 Where your own thinking goes
The same task can use very different amounts of your own effort, depending on how you work with AI.
Own effort by activity (1 = none, 10 = all)
Illustrative model, not measured data
🕵️ Myths and fair points
Tap a question to open it.
Unlikely to be all or nothing. AI is good at some tasks and weak at others, so many jobs will change shape rather than vanish. But some roles and tasks will shrink, and nobody can say exactly how fast. Learning to work with these tools is a sensible hedge either way.
No. Using tools well has always been part of being capable. The real test is whether you can still explain, check and defend the work, because that shows you understood it.
No. Models can state wrong things with total confidence ("hallucinations") and can reflect biases in their training data. This is why a human check on facts and tone matters more as the tools get more fluent.
💡 The C.O.R.E. habit: stay sharp while using AI
- Challenge: ask AI to argue against your idea, play devil's advocate and find the holes.
- Organise: let it handle outlines, formatting and busywork, so your attention goes to the big picture.
- Refine and verify: fact-check every claim. If it uses a concept you don't know, ask it to teach you.
- Elevate: add your own voice, experience and final judgment. AI supplies the bricks; you are the architect.
