Archive · 29 Dec 2025
Professionals Caution on AI-Generated Code Use
A new study found that professional software developers are using AI to help with coding but are careful about how they use it.
They believe AI can help with routine tasks, but important decisions still need human judgment.
Experts warn that AI-generated code can have hidden problems and that relying too much on AI can make it harder to understand and fix systems later.
They suggest that while AI is useful, humans should still be in control and accountable for the code.
The study predicts that issues with AI-generated code, like readability and maintainability, will start to become more apparent in 2027.
Professionals use AI for coding but maintain control to ensure software quality.
AI-generated code can be technically correct but still have subtle issues.
Over-reliance on AI may lead to a loss of deep understanding of systems.
Experts predict a 'Agent Tax' with future issues in readability and maintainability.
AI should be used as a support tool for routine tasks, not for critical decisions.
- Who
- Professional software developers and researchers from UC San Diego and Cornell University
- What
- A study on the use of AI agents for coding and the potential risks involved
- When
- Published on December 29, 2025
- Where
- United States
- Why
- To understand the impact and control mechanisms of AI-generated code in software development
- Study Title
- Professional Software Developers Don’t Vibe, They Control: AI Agent Use for Coding in 2025
- Study Institutions
- UC San Diego and Cornell University
- Publication Date
- December 29, 2025
- CTO of Covasant
- Reddy Raja
- CTO of Vigocare
- Akkiraju Bhattiprolu
- Predicted Diminishing Returns Start Year
- 2027
- AI Code Delivery Year
- 2025