AI Could Be Making Developers Worse, Research Says
▼ Summary
– In February 2026, AI research lab METR attempted to replicate a study measuring developer task completion times with and without AI.
– Developers refused to participate in the study because they would not work without AI, even for a limited number of tasks in a research setting.
– The original study being replicated was conducted in 2025.
By early 2026, researchers at AI lab METR set out to replicate a landmark study measuring how much time developers need to complete tasks with and without AI assistance. They hit an unexpected wall. No one would participate. Developers simply refused to work without AI, even for a limited number of tasks in a controlled research environment.
The original study from 2025 had aimed to quantify the productivity gains from AI coding tools. But the 2026 follow-up revealed something more troubling: a growing dependency on AI that may actually be making developers worse at their jobs. Without the crutch of AI-generated code suggestions, many engineers now struggle to perform basic tasks from scratch.
This trend raises serious questions about skill erosion in the software industry. As developers increasingly rely on AI to write code, debug errors, and even design architecture, their ability to think critically and solve problems independently may be fading. The refusal to participate in the study is itself a data point, suggesting that AI has shifted from being a helpful tool to an essential component of the development workflow.
The implications extend beyond individual competence. If a generation of developers cannot function without AI, the industry could face a productivity paradox: tools designed to accelerate work may instead hollow out the very skills needed to use them effectively. The METR findings serve as a warning that convenience comes at a cost, and that cost may be the expertise of the workforce itself.
(Source: The Next Web)
