Károly Zsolnai-Fehér examines an experiment involving 52 junior software engineersExperimental validation tests a prediction or proposed mechanism with measured observations from a suitable real-world or laboratory experiment. who completed a Python task either with or without an AI coding assistant. The AI-assisted group finished about two minutes, or 8 percent, faster, but that speed difference was not statistically significant.
The learning difference was much clearer. Developers who used the assistant scored 50 percent on a follow-up quiz, compared with 67 percent for those who worked without it, and the largest gap appeared in debugging knowledgeDebugging is the process of finding why software behaved incorrectly and checking that a correction addresses the cause.. Károly Zsolnai-Fehér argues that relying on generated codeCode generation uses AI or another automated system to create source code from instructions, examples, schemas, or higher-level specifications. can leave developers less prepared to diagnose failures themselves.
Károly Zsolnai-Fehér recommends using AI to automate work people already understand, treating it as a tutor when learning something new, and attempting a repair before asking the model for help. The study is limited by its small sample, short task, single library and chat-style assistant, so it is a useful signal rather than a final verdict on AI-assisted programming.
Watch on YouTube
