New research published in leading journals shows that even brief AI assistant use degrades people's problem-solving abilities and critical thinking skills. The findings raise urgent questions about cognitive impacts as AI tools become ubiquitous in education and workplaces.The first real data on what happens to your brain when AI does the thinking. Spoiler: it's not good. This isn't anti-AI scaremongering - it's neuroscience.The study, published in Wired and drawing on research from multiple institutions, found that participants who used AI assistants for just 10 minutes performed significantly worse on subsequent problem-solving tasks compared to control groups. Even more concerning: the effect persisted even after the AI was removed.Researchers tested participants on tasks requiring analytical thinking, creative problem-solving, and basic reasoning. One group had access to AI assistants. The other worked without AI help. Then both groups tackled new problems without any AI assistance.The AI-assisted group performed worse across the board. Not just a little worse - significantly worse. They struggled with tasks they should have been able to handle. They showed less persistence when encountering difficult problems. They were more likely to give up."What we're seeing is a form of cognitive offloading," explains one of the study's authors. "When people know they can ask an AI for help, they don't engage the mental effort to solve problems themselves. And that muscle atrophies quickly."This matches what I've observed in the startup world. Engineers who rely heavily on AI coding assistants often struggle when the AI isn't available. Not because they never learned to code - but because they've stopped exercising the problem-solving muscles that coding requires.The mechanism appears to be straightforward: when you outsource thinking to AI, your brain learns not to think. Use it or lose it applies to cognitive skills just like physical ones. And apparently, you can start losing it in as little as 10 minutes.What makes this particularly concerning is the speed of AI adoption in education. Schools are deploying AI tutors, AI writing assistants, and AI homework helpers at scale. If even brief AI use degrades thinking skills, what happens to students who use AI tools for hours every day?The standard counterargument is that AI is just a tool, like calculators or spell-checkers. We didn't worry about calculators making people bad at math. Why worry about AI making people bad at thinking?The difference is scope. Calculators automate arithmetic. AI assistants automate thinking itself - analysis, reasoning, problem-solving, writing, planning. When you automate the mental processes that make people capable, you risk creating dependency rather than capability.This doesn't mean AI assistants are bad or shouldn't be used. It means we need to be thoughtful about when and how we use them. Using AI to handle routine tasks so you can focus on complex thinking? Probably fine. Using AI to do your thinking for you? Probably not.The research also found that awareness of the problem helps. Participants who were explicitly warned about cognitive offloading effects and encouraged to engage deeply with problems performed better than those who weren't. Simple interventions - like asking people to solve problems themselves before consulting AI - made a difference.This suggests a path forward: treat AI assistants like training wheels. Use them when you're learning or when you genuinely need help. But make a conscious effort to solve problems yourself first. Don't let the easy availability of AI answers stop you from developing your own thinking skills.Education systems should pay attention. If we're deploying AI tutors and homework assistants without teaching students when and how to use them, we might be creating a generation that struggles to think independently. That's not a hypothetical future risk - it's happening now.The workplace implications are equally significant. If employees rely on AI assistants for everything from writing emails to analyzing data, what happens when the AI fails or gives wrong answers? Do people still have the judgment and skills to catch errors?I've seen this in code reviews. Engineers using AI copilots sometimes ship code they don't fully understand. When bugs appear, they struggle to debug because they never developed deep understanding of what the code does. The AI made them productive in the short term and dependent in the long term.None of this means AI assistants are inherently harmful. They're powerful tools that can genuinely help people work more effectively. But like any powerful tool, they have side effects. And we're just starting to understand what those side effects are.The technology is impressive. The question is whether we need it in contexts where the goal is learning and skill development, not just getting answers quickly.Ten minutes. That's all it took in the study. Ten minutes of AI assistance created measurable cognitive effects. We're building a world where AI assistance is always available, always on, always ready to do our thinking for us.Maybe we should pause and ask whether that's actually what we want.