Yes AI can write code way faster than human, but the bottleneck of software development has never been the typing part. When it comes to coding, AI will produce an implementation of a feature and at first glance it will probably look pretty ok. So it feels “quick”, and you move on to the next task and repeat the process and feel more productive than ever.
This productivity is an illusion, you are silently planting time bombs that will blow up in your face way later.
Piling up suboptimal implementations
willthout AI, an experimented developer will treat everything as a problematic that needs to be solved. He will carefully evaluate amongst all the possible solutions he already knows and often comes up with a new ones which fits best the overall context of the existing code and architecture. What AI is doing instead is producing an average (therefore suboptimal) implementation of a feature with poor understanding of the existing codebase around it. Iterating on this, will slowly but surely drift towards a unmaintainable pile of shit.
The slow erosion of the developer skill
Your role is now to review AI code, which is a pernicious shift from an active position to a passive one. A good developer never stops learning and sharpens his skills every time he produces code by himself. While validating and tweaking AI code instead of producing his own, the developer will at best stagnate and most probably on the long run begin to erode his development skills. The “augmented developer” will end as a “retarded developer” at some point.
Conclusion
The whole illusion of AI productivity stands there, on the visible short term “gains”. There are no shortcuts when it comes to software development, only tradeoffs. I’ve never had a single line of my code written by an AI, and I probably never will.
