AI Coding Tools: Surprising Study Challenges Developer Productivity Claims

Key Points

1. The METR study involved 16 experienced open-source developers, with half using AI tools like Cursor Pro and the other half working without AI assistance.

2. Contrary to expectations, developers using AI tools took 19% longer to complete tasks, despite predicting a 24% reduction in completion time.

3. The study suggests that AI tools, particularly 'vibe coders', may slow down experienced developers due to context and prompt-based code generation.

4. The authors caution against broad conclusions, noting that AI's rapid advancements could change its impact on productivity in the future.

5. Other studies indicate that AI-generated code often requires additional review and testing, adding complexity to workflows.