45c3IF85N
Tinker ☀️RE: https://social.coop/@cwebber/116217717944115087
Systemd (popular component to many {most?} Linux distributions) getting enshittified by AI.
Read @cwebber 's thread quoted below for more.
Also, thanks a lot (thanks a bot?) daandemeyer.
Francesco P Lovergine 
@tinker @cwebber
I wonder how a context-limited AI tool can pretend to perform automagically a code review successfully. Someone has not clear in mind what the limits of such tools are. At least I can appreciate it is read-only, but a lot of people are not so crafty and noise could always be more than signal.
I wonder how a context-limited AI tool can pretend to perform automagically a code review successfully. Someone has not clear in mind what the limits of such tools are. At least I can appreciate it is read-only, but a lot of people are not so crafty and noise could always be more than signal.
catraxx 
Anarchozionist/Datenprolet
FandaSin
MegatronicThronBanks@megatronicthronbanks - Right?
degenerating degenerateAnyone who objects could approach the maintainer and offer to do the work manually instead... ?
@tinker So if he does use Claude to manage issues and suchlike, that's bad, because... ?
Github has offered bots and such for the same kind of thing for many years.
I noticed there's some rhetorical sleight of hand here: coding AIs do not simply produce "slop". What they produce nowadays is generally good code.
Nor is using AI == "enshittification", I am betting it makes this project better much more than it makes problems (and humans also make problems...)
@hopeless - Something bad happening for long enough does not make adding the bad to it okay. Github offering similar for "many years" makes this worse, not better or a comfort.
The use of AI to review vs the use of AI to contribute (either directly or through a proxy) isnt much different and the project as a whole has indicated wider acceptance of AI.
So far as the use of AI being problematic, those are well documented and I question your presented naivety and sincerity.
- Hallucinations (even agents with "good output" make these mistakes) and the likelihood of introducing vulnerable code
--- Arguments to "they can QA it" competing with the higher volume of shit output that needs to be QA'd
- Tertiary costs of higher water usage and energy costs
- Tertiary costs of high pollutants
Etc etc.
"that's bad, because" - you can disagree with the because, but please dont feign ignorance.
I mention github having bots for many years to point out that the sky didn't fall.
Using AI to review issues is quite different from code; but since I think it would be helpful in both cases, no need to argue about it.
"AI being problematic" is just the usual dark mutterings. Today's coding AI is clearly very useful since vast numbers of people are using it.
The guy using Claude is bad because [stock arguments about any AI in Mar 2026 when out of ideas] doesn't convince, sorry.
@tinker FYI
"Google Cuts Power Needs for AI Queries 33x in One Year"
https://techreport.com/news/software/google-ai-energy-efficiency/
This happened leading up to Gemini 3.0 era (Aug 2025)
@hopeless - Yeah, we're going to be talking around each other.
I'm glad this very specific instance is working for you. And I'm glad you don't have to deal with any negative externalities or knock-on effects.
Some of us don't have that luxury.
But again, I think we're arguing from two different areas, goals, and consequences.
I hope you have a good weekend.
KielKontrovers BlogYes I also get stupid stories about the scary AI here on Masto every day. However I am also having generally big successes on my actual code with the coding AI every day.
So especially when people I link to antigravity will run a mile rather than try it themselves, I tend to think they don't know what they're talking about, but instead bobbing along behind the loudest, most strident voices.
Harald Hannelius 
