I wasn’t at first until I saw all the abandoned vibe coded projects. Now I have to spend extra time trying to figure out which projects are being vibe coded by people who don’t know what they are doing and which projects have competent devs that can read, and understand the code that’s being spit out.
I couldn’t agree more. Four years ago, I would have been very excited to see a new FOSS project come out. Now I find myself immediately scanning their repo for the usual suspects (AGENTS.md, CLAUDE.md, LLMs as contributors), maybe check their release cadence and lines of code per commit, and then decide to come back in a year or two to see if the thing is still alive at all.
That’s actually my reasoning too. This is someone with a claude subscription who thinks they can do a better job than libre office, and it will probably get shelved the second they get bored. To me that’s arrogance.
Vibe coding is fine for fiddly things you do yourself, but don’t try to say you’re going to do a better job than libre office.
I agree. I’ve used a local LLM to vibecode bash scripts for automating my repetitive and annoying tasks. I can get the gist of what is happening in the scripts but I can’t fully review them. If one of these scripts deletes all my work that’s my own problem - no way would I expect other people to run them.
This has become really annoying and problematic. Particularly with small and niche software like plugins that don’t get called out anywhere.
It has kind of broken my trust in developers. Now, when I install something that doesn’t work I don’t want to risk wasting my time filing a bug report because it feels likely to be a lost cause. I just move on and become a little bit more cynical.
I just wish that developers would be more honest and explicit about their level of understanding of what they’re publishing and how LLMs have been used. I’ve locally vibe coded some bash scripts that other people could find useful, but I wouldn’t want anyone else to run them unless they were fully aware that I don’t completely understand the code. Kinda feels like common decency tbh!
I wasn’t at first until I saw all the abandoned vibe coded projects. Now I have to spend extra time trying to figure out which projects are being vibe coded by people who don’t know what they are doing and which projects have competent devs that can read, and understand the code that’s being spit out.
I couldn’t agree more. Four years ago, I would have been very excited to see a new FOSS project come out. Now I find myself immediately scanning their repo for the usual suspects (AGENTS.md, CLAUDE.md, LLMs as contributors), maybe check their release cadence and lines of code per commit, and then decide to come back in a year or two to see if the thing is still alive at all.
This. The sheer volume is an objective problem. It’s basically digital litter.
That’s actually my reasoning too. This is someone with a claude subscription who thinks they can do a better job than libre office, and it will probably get shelved the second they get bored. To me that’s arrogance.
Vibe coding is fine for fiddly things you do yourself, but don’t try to say you’re going to do a better job than libre office.
I agree. I’ve used a local LLM to vibecode bash scripts for automating my repetitive and annoying tasks. I can get the gist of what is happening in the scripts but I can’t fully review them. If one of these scripts deletes all my work that’s my own problem - no way would I expect other people to run them.
This has become really annoying and problematic. Particularly with small and niche software like plugins that don’t get called out anywhere.
It has kind of broken my trust in developers. Now, when I install something that doesn’t work I don’t want to risk wasting my time filing a bug report because it feels likely to be a lost cause. I just move on and become a little bit more cynical.
I just wish that developers would be more honest and explicit about their level of understanding of what they’re publishing and how LLMs have been used. I’ve locally vibe coded some bash scripts that other people could find useful, but I wouldn’t want anyone else to run them unless they were fully aware that I don’t completely understand the code. Kinda feels like common decency tbh!