AI Slows Down Experienced Developers: A New Study’s Surprising Discovery

AI Slows Down Experienced Developers: A New Study’s Surprising Discovery

Ondřej Barták
Ondřej Barták
Entrepreneur and Programmer
15. 7. 2025
2 minutes reading · 13 views
AI Slows Down Experienced Developers: A New Study’s Surprising Discovery

AI Slows Down Experienced Developers: A Surprising Finding from a New Study

A new study conducted in early 2025 produced a surprising finding—artificial intelligence actually slows down experienced open-source developers instead of speeding them up. The research tracked 246 tasks completed by experienced developers on large, popular repositories and found that using AI tools increased implementation time by 19%.

Expectations Versus Reality

Before the experiment began, the developers predicted that AI would help them reduce implementation time by 24%. Even after the study was completed, the developers continued to believe that AI had made them 20% faster, despite the reality being the opposite. Experts in economics and machine learning were even less accurate in their predictions, forecasting a 38–39% speedup.

How Developers Spend Their Time with AI

An analysis of 74 screen recordings revealed interesting changes in how developers spend their time. When using AI, they:

  • Spent less time actively programming and searching for information
  • Devoted more time to communicating with AI systems, waiting for responses, and reviewing results
  • Surprisingly, also spent more time being inactive

Why AI Slows Developers Down

The researchers identified several key factors contributing to the slowdown:

Low AI reliability: Developers accepted fewer than 44% of suggestions from AI systems, and most reported having to make significant changes to AI-generated code.

Extensive repository knowledge: Experienced developers, with an average of 5 years of experience and 1,500 commits to the repositories in question, often knew better than AI how to solve the problems.

Complex environment: Large repositories averaging more than 1,100,000 lines of code posed a challenge for AI systems, which struggled with context.

Future Outlook

The study also showed that autonomous AI agents using Claude 3.7 Sonnet can often implement the basic functionality required by tasks, although they fail on important details such as documentation or testing. This suggests that the situation may change in the future as AI technologies continue to develop.

The results of this study highlight the importance of real-world experiments rather than relying solely on expert predictions or developer surveys. They also show that context and experience play a key role in the effectiveness of AI tools.

Advertisement

Content created with help from UpTier.

SEO and GEO on autopilot. UpTier’s multi-agent systems write and optimize content for search engines and AI answers.

Discover UpTier ↗

Category:AI
Did you enjoy this article?
Discover more interesting posts on our blog
Back to blog

Related posts

Google makes Tasks free for Gemini 4 Hydrogen usersGoogle makes Tasks free for Gemini 4 Hydrogen users
Tasks is becoming available for free to hundreds of millions of Gemini 4 Hydrogen users. Google says it can call businesses, book appointments and complete purchases on their behalf.
1 min read
18. 11. 2026
OpenAI introduces GPT-6.2 Astra and Sol for professional and everyday workOpenAI introduces GPT-6.2 Astra and Sol for professional and everyday work
OpenAI has introduced two models: Astra for complex professional tasks and Sol for everyday work. Both are available in ChatGPT Work, Codex and through the API.
3 min read
17. 11. 2026
Tencent releases Hy4 Flash with image support and a context window of up to 1 million tokensTencent releases Hy4 Flash with image support and a context window of up to 1 million tokens
The lighter Hy4 Flash model targets chat, documents and routine code edits. It accepts text and images, supports tool use and, according to Tencent, matched Claude Haiku 5.5 in tests of everyday tasks.
1 min read
16. 11. 2026
Přihlaste se k odběru našeho newsletteru
Zůstaňte informováni o nejnovějších příspěvcích, exkluzivních nabídkách, a aktualizacích.
CodedTrip

Operated by CodedTrip LLC, USA.

YouTube
TikTok