zlacker

So then we'd have 100x as much "vibe coding" as now? <shudders>

Even if I disagree with our AI code assistant's usefulness - it often just suggests code based on "someone else in a different file used this statement after that last API call" that is totally irrelevant to my current needs - I can usually assess AI-generated code (C++ / python) relatively quickly and determine if it's what I want or not. It generally nails the "you just created this variable and now want to print it to the log" part pretty well.

Debugging AI-generated perl?? It would be like trying to follow "obfuscated C contest" entries all day, every day. No thanks!

>>hn_acc+(OP)
The flip side to this is that perl often favored brevity over simplicity. It turns out that this is also a really useful attribute when working with LLMs. I'm not sure how confident I am here, but it really wouldn't surprise me if LLMs could both generate and explain remarkably complex perl code!