The AI Productivity Trap

For me, whether AI can be beneficial or not is out of the question, and I am not here to discuss that in much detail. I already heavily use tools like aider (a terminal-based alternative to Cursor), autocompletion assistants such as copilot, and conversational LLMs like claude, chatgpt, and deepseek all the time. So, I truly believe there’s immense value these tools can provide, and I’m not here to discourage anyone from using them. They’ve already proven their worth to me in many ways.

However, the thing is, no matter how much AI increases our productivity, at the end of the day, that productivity increase is a multiplier of our existing productivity. The key determining factor remains how well we have mastered the basics and our productivity level without AI, not with it. This is also partly why I believe that existing AI tools are much more useful for people who are already experienced in a field than for newcomers.

The Illusion of Improvement

When someone first starts using AI and experiences a productivity boost, they might find themselves completing more tasks in less time than before. This can create the illusion that they have improved. But in reality, the multiplier effect on their base productivity can also lead to a decline in their capabilities. Yes, you can achieve more results in the same amount of time, but this also means you can achieve the same results with less effort. Less effort means less time spent on self-improvement. And humans have a natural tendency to spend as little of energy as possible.

A fitting analogy would be the invention of cars. The invention of cars allowed us to travel greater distances. But since we could travel greater distances with less effort, they also have led to muscle atrophy for many, as they no longer needed to rely on their leg muscles. I strongly suspect that over-dependence on AI could lead to similar consequences.

The fact that some new graduates struggle to write code without LLM support is a clear example of this phenomenon. How these tools impact our learning is a very important topic to further discuss. If the short-term productivity gains is harming the learning process, I would rather not use AI at all and rely on my own skills than relying on AI, which is yet another external dependency.

Hopefully, it should still be possible to benefit from the good sides of AI while keeping the side effects in control. Just as using cars does not prevent us from still working out our legs.

Make sure you’re not relying on AI so much that you’re not just outsourcing tasks but also the process of learning. Occasionally write code without using AI tools to keep your skills sharp. Be cautious about the suggestions AI tools give you. I’ve found AI more helpful for tasks I already understand well but don’t want to do myself. However, when I use it in areas I’m not familiar with, the code often becomes messy. The quality of the answers you get depends on the quality of your prompts, which depends on how well you know the topic. So, always give more priority to actually learning the concepts that you are working on top off.

Conclusion

The main idea to take away is that AI mostly acts as a multiplier on your existing productivity and that you have to prioritize your base productivity levels rather than your AI-enhanced productivity levels.