{"id":7303,"date":"2025-08-18T12:44:36","date_gmt":"2025-08-18T19:44:36","guid":{"rendered":"https:\/\/www.ultimatewb.com\/blog\/?p=7303"},"modified":"2025-08-18T12:45:58","modified_gmt":"2025-08-18T19:45:58","slug":"ai-coding-tools-make-developers-slower-even-when-they-think-theyre-faster","status":"publish","type":"post","link":"https:\/\/www.ultimatewb.com\/blog\/7303\/ai-coding-tools-make-developers-slower-even-when-they-think-theyre-faster\/","title":{"rendered":"AI Coding Tools Make Developers Slower – Even When They Think They\u2019re Faster"},"content":{"rendered":"\n
\"Coding<\/a><\/figure>\n\n\n\n

Artificial intelligence is supposed to speed up software development, but a new study suggests that the reality is more complicated. Researchers from Model Evaluation & Threat Research (METR)<\/strong> conducted a controlled trial and discovered that AI coding tools can actually slow developers down – even when the developers believe they\u2019re working faster.<\/p>\n\n\n\n

The Study: Expectations vs. Reality<\/strong><\/h2>\n\n\n\n

The METR study involved 16 experienced developers<\/strong> working on large open-source projects. They tackled 246 real-world tasks<\/strong>, ranging from bug fixes to new feature implementation, with some tasks allowing the use of AI tools and others not. Developers were free to choose AI tools like Cursor Pro<\/strong> paired with Claude 3.5\/3.7 Sonnet<\/strong> when allowed.<\/p>\n\n\n\n

Before starting, developers predicted that AI assistance would boost their speed by 24%<\/strong>. After finishing the tasks, they reported feeling 20% faster<\/strong>. In reality, AI use slowed task completion by 19%<\/strong>.<\/p>\n\n\n\n

\n

\u201cSurprisingly, we find that allowing AI actually increases completion time by 19 percent – AI tooling slowed developers down,\u201d the study notes.<\/p>\n<\/blockquote>\n\n\n\n

Why AI Tools Can Hinder Productivity<\/strong><\/h2>\n\n\n\n

The researchers identified five main factors<\/strong> contributing to the slowdown:<\/p>\n\n\n\n

    \n
  1. Over-optimism about AI usefulness<\/strong> – Developers expected AI to save more time than it realistically could.<\/li>\n\n\n\n
  2. High developer familiarity with repositories<\/strong> – Experienced developers already knew the code well, leaving little for AI to add.<\/li>\n\n\n\n
  3. Large and complex repositories<\/strong> – AI struggles with massive projects, especially those exceeding a million lines of code.<\/li>\n\n\n\n
  4. Low AI reliability<\/strong> – Less than 44% of AI suggestions were accepted, forcing developers to spend time reviewing and correcting output.<\/li>\n\n\n\n
  5. Implicit repository context<\/strong> – AI couldn\u2019t fully grasp the nuances of the projects, limiting its effectiveness.<\/li>\n<\/ol>\n\n\n\n

    Other factors like latency in AI generation and suboptimal prompts may also have played a role, though their impact is unclear.<\/p>\n\n\n\n

    The Hidden Work of AI<\/strong><\/h2>\n\n\n\n

    One key insight from the study is that AI doesn\u2019t reduce work so much as shift it<\/strong>. Developers spent less time actively coding and more time:<\/p>\n\n\n\n