Unintended consequences of GenAI in software projects, so far

Interpretations and intuitions based on the available research. So what’s this all about?

How AI-assisted coding affects developers and projects

Check out the links to research are at the end of the article.

‘Faster’ coding

Research on the impact of AI tools like GitHub Copilot on software projects is a double-edged sword. It claims developers are writing code 55% faster with Copilot, which presumably improves ‘developer productivity’ (whatever it exactly is, let’s leave the discussion about it for some other time).

BUT. This increase in speed comes with potential downsides regarding code quality and maintainability. This study shows a trend of code being revised and rewritten twice as often compared to the pre-GenAI era. This is as if whole teams are becoming fast code-producing but more-junior developers.

Not DRY (Don’t Repeat Yourself)

So far it seems that the motivation of the GenAI is to produce code that would deliver on the needed functionality, rather than code that fits with the bigger picture, is reusable and legible. Plus, it appears that there are often lines of code suggested that are not needed there in the…