A recent study shows that developers using AI tools actually slow down. The reason might be deeper than the tools themselves.
A new study from METR caught my eye this week. It looked at open-source developers using AI tools to complete GitHub issues. The result? Developers with access to AI took 19% longer to complete tasks.
That's surprising enough. But here's the twist: developers expected AI to make them 24% faster, and still believed it had sped them up by 20%, even after it made them slower.
That gap between perception and reality says something important, and I think it extends beyond developers. It's not that AI isn't useful, it clearly is. But in some cases, it's starting to hinder more than it helps. We've begun to doubt ourselves. People now second-guess things they would have once done confidently. Rather than act on instinct, knowledge, or skill, they reach for AI. And when the output isn't quite right, they keep tweaking the prompt instead of trusting their own judgment.
AI has become a crutch. A shiny one, yes, but one that subtly chips away at speed, confidence, and autonomy.
This is why the future of AI tools won't revolve around prompt boxes and retries. The next generation will be invisible specialists, not generalist sidekicks. These systems won't wait for the user to ask. They'll act on deep context, executing specific, high-skill tasks in the background without needing constant supervision or correction.
We won't need to ask them to write the thing, or fix the thing, or explain the thing. They'll already know. And they'll just do it.
AI that works like this won't slow us down. It will get out of our way and make us faster, better, and more confident than before.