A recent study involving experienced developers found that, contrary to expectations, AI increased task completion times by 19%, mainly due to the time spent reviewing and correcting AI suggestions.

The authors of the study remain optimistic, suggesting that these findings do not reflect a fundamental limitation of AI and that future improvements could enhance AI's effectiveness.

Most participants and researchers continue to use AI tools like Cursor because they make development less effortful, even if they don't necessarily speed up the process, and the impact may vary depending on the developer's experience and the task complexity.

Future research is planned to continue monitoring AI's impact on software development, as understanding these dynamics is crucial for AI R&D and managing associated risks.

This slowdown challenges the common belief that AI consistently boosts productivity among seasoned engineers, despite significant investments in AI development tools.

Despite these concerns, not all layoffs are directly replaced by AI, indicating a complex transition in the workforce rather than straightforward automation.

A recent study by AI benchmarking nonprofit METR, focusing on tools like Cursor Pro, revealed unexpected results regarding AI's effectiveness in coding assistance.

The study identified five factors contributing to the productivity decline and confirmed that the results were not due to experimental artifacts, although the methodology lacked a blinding element.

Meanwhile, industry leaders like Dario Amodei of Anthropic warn that AI could eliminate up to half of all entry-level coding jobs within the next one to five years, raising concerns about widespread job displacement.

Previous research had reported significant productivity gains from AI, but the latest study's surprising results highlight the need for realistic evaluations of AI capabilities.

The lead authors expressed surprise at the findings, which contradict earlier optimistic reports, and noted that AI tools seem more effective for less experienced developers or simpler tasks.

Factors such as overuse of AI, developers' lack of experience, and high standards for open-source projects contributed to the observed slowdown, emphasizing that AI's impact varies widely.