Here’s to the AI contrarians.

The more capable AI becomes, the more deliberately I want to practise the things I value: reading, writing, reasoning and making things.

There are people taking a similar approach. They use AI to explore an idea, organise work or get feedback. They also keep a book open, reach for a notebook and spend time developing a craft. I call that an AI contrarian mindset.

Speed is useful. So is being able to explain how you reached an answer. When AI removes some of the effort from producing work, we can choose where to put that effort back.

Keep the useful difficulty.

A book asks you to follow an argument beyond the headline. A blank page asks what you actually think. A practical skill gives you feedback that a polished paragraph cannot: the joint does not fit, the rhythm slips, the drawing feels wrong.

Choose difficulty that develops something you care about. You can use AI for hints and feedback while still doing the attempt yourself. Paper is one option; accessible writing, audio and other tools can serve the same purpose.

For example, build the budget before asking for feedback.

You are planning a small shop. List the costs and explain which ones change with sales. Sketch the break-even logic yourself. Then ask AI to challenge the missing costs and assumptions. Check the calculations in a spreadsheet and explain why the business would or would not cover its costs.

You finish with a better budget and a clearer understanding of it. The quality still depends on the inputs, checks and decisions you make.

Use the research carefully.

A 2025 survey of 319 knowledge workers found that higher confidence in generative AI was associated with less self-reported critical-thinking effort. It studied reported experience, so it cannot establish that AI caused a loss of ability. Lee and colleagues, CHI 2025.

In experiments with science texts, students who practised recalling and reconstructing what they learned performed better on later conceptual tests than students using elaborative concept mapping. That supports trying to explain an idea from memory and then checking it. It does not establish that every difficult activity improves general intelligence. Karpicke and Blunt, 2011.

Put the saved time somewhere worthwhile.

Let AI remove work that adds little value. Keep a regular place for sustained reading, a first draft in your own words and practice in a skill you want to develop. You decide what deserves the effort.