Hey HN,
There is an interesting ongoing tension in the discourse around the quality of AI-generated code and whether "coding is solved".
As an experienced but very average programmer, the powers that AI have unlocked for me are nothing short of incredible. At work I am far more productive, and personally I have created/generated software that I would never have been able to create. For example, I've created my own very functional DAW (Digital Audio Workstation) that has tight integration with specific hardware I own.
I'm wondering if we are past the point of "AI generates bad and unmaintainable code". My gut tells me that for the most part we are, and my reasoning for this:
A) AI agents are able to find and fix many issues that I would never have caught B) They are FAR faster than me C) They generate WAY more tests (admittedly, sometimes it's out of hand) D) If something starts to really go sidewise in terms of code organization, you can often use AI to do the refactoring and cleanup and re-testing while you are out for a walk or having dinner with your family...
We are still fairly early into this experiment though. Do you fear that AI is generating unmaintainable code? Is the industry going to have major problems in the future with technical debt that more use of agents can't get us out of?
Do you have tips for how to coach agents into creating code that is correct and maintainable long term?
Comments URL: https://news.ycombinator.com/item?id=50508896
Points: 2
# Comments: 0
