Why does AI help today but leave you weaker tomorrow?
Your team ships faster with AI, then stalls the moment a problem falls outside the tool's reach. A new study suggests the cost is not time saved or spent, it is persistence itself.
The experiment that should worry you
Cal Newport covers a paper titled "AI Assistance Reduces Persistence and Hurts Independent Performance," presented at the Conference on Language Modeling. Participants solved fraction math problems, one group alone and one with ChatGPT. The ChatGPT group scored better at first, which is expected. Then researchers turned the chatbot off, and the AI group's accuracy fell below the control group's. Worse, many in that group gave up and started skipping questions entirely.
Why the drop happens
Per Wikipedia, sustained attention is the voluntary maintenance of goal-directed focus over time. It relies on executive function, the brain's ability to plan and resist distraction, plus working memory. These are the exact capacities you skip when an AI hands you the answer. The paper's title names the second cost: persistence. If you rarely practice pushing through a hard problem unaided, quitting becomes the default response to friction.
What deep focus actually buys
Wikipedia describes flow as full immersion in an activity, with energized focus and enjoyment in the process. Flow depends on sustained attention holding long enough for immersion to happen. Constant tool handoffs break that state before it starts. So the real trade is not speed versus slowness, it is output today versus trained concentration capacity tomorrow.
How to buy the capacity back
Newport frames hard things as good precisely because they are hard: effort is the training signal. A practical split follows. Use AI for tasks you never need to master, and go unaided for skills your business depends on. Alternate deliberately, so the assisted mode does not silently become the only mode.
What this does not solve
One study on math and reading tasks does not prove AI erodes all cognition. It does not measure experts, long timelines, or creative work. And it cannot tell you where your own team's line between use and dependence sits. Treat it as a warning signal to test, not a verdict.
Which skill in your team is so core that going unaided should stay the default?
How would you detect the moment assisted work quietly became the only mode?
Pick one recurring task this week and do it without AI, then note where you wanted to quit.
Sources
Recommended for you
- Durable Skills
Why does your fast learning of a new domain still not stick?
You read, you cram, you forget. Metacognition, spacing, and transfer explain what fast learning actually requires.
- Durable Skills
Why does a confident AI answer still leave you unsure?
When a machine states something fluently, your own biases decide whether you push back. Three old ideas explain why that fails.
- Durable Skills
What breaks when you treat code review as bug catching?
Reviewers who only hunt for bugs miss the design and complexity problems that cost you later. Google's own guide explains why.
- Durable Skills
What breaks when your agents run for days?
Agents now run for hours or days, but most orchestration frameworks were built for tasks that finish in minutes. Nobody owns the gap.
Enjoyed this article?
Subscribe for new articles. No spam. Unsubscribe anytime.