If genAI were a human employee it would have been fired years ago for gross incompetence. Despite on the glitzy announcements of "improvements", it's not getting any better. Really makes you wonder where all those billions of dollars are actually going.
I’m sure this runs the gamut, but I think vibe coding, when effective, I find to be an unabashed win, it was always about the output for me, and making mind to page faster for me is good. But then again, I’m in devops, not software engineering. I’ve done more personal projects than the mean devops engineer, but one thing I’ve noticed is: streamlining the process of making games, workflows, and my favorite thing, dashboards is just all positive. My only real complaint about vibe coding is that it’s not where I want it to be. It can’t keep up, often doesn’t know what I want, abides and patterns and best practices that are in a nuanced way different than the problem I’m trying to solve, and worst of all, constant regressions, particularly when trying to link different vibe coded modules together.
In some sense it makes me wonder if this is a solve able problem. Can you have an AI optimize both for understanding previous patterns and for innovation? Can you really dial in just the right amount of autocomplete recommendations? When it is common to have a part of your code that goes against the patterns in other parts of your code? Right now, there are so many autocompletes in cursor that just distract me from my coding, and I have to intentionally ignore them and not accidentally press tab, because the pattern I started is not the pattern I want to continue.
"Those resisting vibecoding are delaying the inevitable." The inevitable being fragile codebases with so much tech debt it will make maintaining an application from the 80s seem like a walk in the park? If that's the inevitability you're talking about I'll gladly skip it, thanks lol.