I'm really not seeing a lot of code that I can say is bad AI code.

I and my coworkers use AI, but the incoming code seems pretty ok. But my view is just my current small employer.

The breadth of industry is so vast that people have wildly different takes on this. For a lot of simple coding tasks (eg custom plugins or apps) an LLM is not only efficient, but extremely competent. Some traditional coders are having a harder time working with them when a major challenge comes from defining the problem and constraints well. It’s usually something kept in head. So, new skill sets are emerging and being refined. The ones that thrive here will not be coders, but it will be generalists with excellent management and communication skills.

	▲	duxup 8 hours ago \| parent [-]
		Yeah most of my team is using an LLM for "make this function better", or learning, or just somewhat smaller bites or code that an LLM will work well with. So we don't see the "hey rewrite this whole 20 year old complicated application, omg it didn't work" kind of sitiatons.

▲

dgunay 6 hours ago | parent | prev [-]

Most of the AI generated code I review is pretty much okay. Usually does what it should and meets some standard of quality. But it usually looks and feels just slightly stylistically foreign compared to the code around it. I personally edit mine before code review so that it looks how I would have written it, but there are many chunks of code in our codebase now where the author and reviewer didn't do that.