Depends on the type of programming. The output is still inconsistent and a little sloppy from real-life examples I’ve seen, and my personal attempts at getting anything useful have been fruitless, but I’m not doing boilerplate junk.
Yeah, I tried just to be lazy and generate some regex and config file for a custom log format in lnav (multi log file viewer - at least for Linux, don’t know about other platforms)
And it pretty much ran circles with the mistakes it made. Every time I corrected it with something, it would make a mistake from before again.
In the end I did the thing myself and could have spared me off nearly an hour of trying this stupid trend of vibe coding…
I’ve found humans to not vary their style so wildly from module to module. Humans also make human mistakes that are of a different nature from LLM hallucinations.
Sure, but we’re still not even close to useful AI agents. We need ones that are correct, not just designed to sound correct.
For programming we are well into useful territory, but definitely not for other walks of life
Depends on the type of programming. The output is still inconsistent and a little sloppy from real-life examples I’ve seen, and my personal attempts at getting anything useful have been fruitless, but I’m not doing boilerplate junk.
Yeah, I tried just to be lazy and generate some regex and config file for a custom log format in lnav (multi log file viewer - at least for Linux, don’t know about other platforms)
And it pretty much ran circles with the mistakes it made. Every time I corrected it with something, it would make a mistake from before again.
In the end I did the thing myself and could have spared me off nearly an hour of trying this stupid trend of vibe coding…
I’ve used it extensively, and yes it can have inconsistent output but so can humans, useful doesn’t imply perfect imo
I’ve found humans to not vary their style so wildly from module to module. Humans also make human mistakes that are of a different nature from LLM hallucinations.