So my systems recently updated to rsync 3.4.3, and as soon as that happened my backup system - which does incremental backups using multiple --compare-dest= arguments - started to fail on anything but a full backup.
Revert to 3.4.1 and it works.
So I go look at the source in GitHub to see what might have changed, because there doesn't seem to be anything relevant in the changelog.
Since 3.4.1, 36 commits by "tridge and claude"
Oh for fuck's sakes.
Leah, you're far more active in the Voidlinux than I am; what's the state of Voidlinux regarding upstream AI slop? Any chance for there to be a void-repo-nonslop, or rather a void-repo-sloppy ?
@mav
@JeremiahFieldhaven
Been a while since LLM/AI was shoved into ArduPilot, which is a safety-critical sUAS firmware.
I have not taken my stuff out since I saw that land in the project.
Edit:
That was the only context in which I knew of tridge and his storied reputation for making safety-critical code.
@chris @sinbad @JeremiahFieldhaven Here's my question... is it really, solving the problem for you? Like, actually? Given all of the costs in the full context of how it operates?
I don't believe that it is.
Can you trust everything it outputs? Are you able to catch any problems with it 100% of the time? Are you somehow able to avoid it anchoring your thinking around a particular method?
Let's assume it does solve the problem, and that somehow a purely ethical AI is produced that magically solves the labor, environment, plagiarism issues, and that is correct 100% of the time.
Even if it does, you are slowly eroding your ability to solve problems of that nature independent of the agent.
No matter how careful you are, no matter how smart, how skilled, how well-versed.
You cannot beat cognitive surrender.
@chris @sinbad @JeremiahFieldhaven So, given that we have NOT solved all those other problems, do you think there's even a 10% efficiency gain?
I think I was pretty generous with 10%.
@chris @distractal @sinbad @JeremiahFieldhaven i dont think you know what those words mean
and if your solveing depression with AI well thats depressing and sounds like how people talk about alcohol
@glassresistor Nope! I don't know in the slightest what it means, and I can't guess either. I also have to consult a dictionary to define pretty much anything in a reasonable fashion, so I guess I have no firm grasp of the English language, either.
I bet like most of the people I've run into on this network of servers, that you probably have a doctorate on the subject, though. Everyone I know seems to have a doctorate on something, while I, the lowly slacker, never aspired to finish public school and took the GED to get the hell out of structured education. Discipline is not for me. More like being constantly disciplined.
@chris @distractal @sinbad @JeremiahFieldhaven curious how close to addiction itll feel like at 10x the cost
and my dude outside of $ loss if your gaming like an addict its an addiction
I'd chime in here and say that good people suffer from substance abuse. Living in Las Vegas and watching the penny pinchers succumb to a big payout from a slot machine was both fascinating and very sad. AI gives a programmer the illusion that they are now the power code maker they always envisioned themselves to be. That isn't much different than cocaine. We're losing good people to this.
@blausand You have put the strikethrough in the wrong place, but more to the point, no, I think that's wrong. I've yet to see any solid evidence it makes anyone smarter under any circumstances.
They say or maybe even *think* they're smarter, yes, but that is *not* the same thing.
This is all about appearance versus fact, and that's what you have muddled up.
@chris Just like I'm sure that saying cognitive surrender is not something a person can "fight" and saying that fighting depression with a chatbot is likely to make the depression worse in the long term is the same as comparing using a computer to gambling.
Jesus Christ, get over yourself.
@trantion @distractal @sinbad @JeremiahFieldhaven I don't always hand it all over to an AI. If it looks like an interesting task I can fulfill easily, I will do it on my own. It's just that I've fallen into the trap of asking prompts to solve what seem like "quick" problems. Unfortunately, some of them are not so quick, since projects don't agree with the reports I used to draft the solution.
For example, a bug currently affecting Wine Wayland under Mutter, Gnome's compositor. Somehow Wine Wayland is abusing a not-well-defined behavior of the protocol that ends up waiting on FIFO protocol on unmapped windows, and both KWin and wlroots happen to dodge this bullet by ignoring FIFO on unmapped surfaces already. Mutter doesn't want to do this the same way, and the developers aren't being terribly forthcoming about what I should do about this. My last attempt at a solution turned out to be wrong according to them, and the second attempt, I chose poorly for the solution strategy again.
I should just let the two sides fight it out, except this unmitigated issue results in Proton CachyOS simply hanging on startup half the time, and only on Gnome/Mutter. I guess I can just carry my workaround locally and be the only person who benefits from the incorrect solution.
If the choice is between ruining something important with LLM slop or not doing anything, the latter is the correct choice. If you want to make yourself stupider that's your choice but the moment you ruin a public project with it, that's a different story. The public is not your therapist.
@distractal @chris @sinbad @JeremiahFieldhaven
I am sharing a critical view regarding 'cognitive surrender', yet a reality check on expectations:
"... a purely ethical AI is produced that magically solves the labor, environment, plagiarism issues, and that is correct 100% of the time."
=> does code written by humans fulfill all these criteria, I mean even approximately?
@chris Your understanding of the code base was lacking, so you spent 5-15 minutes making it worse. Is that really the way to go?
@chris @distractal @sinbad @JeremiahFieldhaven
Its a common misconception that LLMs can solve problems. They are, fundamentally, pattern matching systems. I have to use these at work and review code that is co-authored by a model. Ime at least, producing slop is a human error caused by misunderstanding what LLMs are and what they can actually do.
This is why they are so good at picking up on security vulnerabilities. They look for permutations of known attacks by identifying patterns. Some trivial fixes can be auto corrected, but a human always has to be in the loop to make any actual judgement calls. LLMs work in probability, not correctness.
I got bills to pay and my job is knowing how shit works. There is no ethical consumption under capitalism 
@chris @distractal @sinbad @JeremiahFieldhaven
"So my systems... started to fail on anything but a full backup."
that doesn't sound like "solve a problem for me in 5-15 minutes" That doesn't sound like solving a problem at all.
(removed larger list of accounts not related to this subthread)
@TurboTrain Well 😒 ...
"Oh, No! Not #OpenZFS?!"
"Et tu, OpenZFS?"
"Slop comes for every project."
@[email protected] With https://mastodon.social/@bms48/116662724883486066 in consideration, you are now moving goal posts, or accepting the increasing use bit by bit -- pick as suitable.
Oriel Jutty 
Jeremiah Fieldhaven
Leah Neukirchen
datenwolf
mirabilos🐈⬛
mav 
Daniel Gibson
Brett Edmond Carlock
The Seven Voyages Of Steve
Erik Johnson
Christopher Snowhill
💥meta physical deflationist💥
Maya
Chuck
blausand 🐟
Liam Proven
Vincent Sparks
David Gerard
Trantion
gsec
chris 🔓 Digital Sovereignty
Orb 2069
krig
jgeorge
Annika Backstrom
chx
Dr. Bruce Simpson
EndlessMason
𐑮𐑨𐑛𐑦𐑒𐑩 ·𐑦𐑒𐑕𐑡𐑱
goedelchen
sosquqer
ltning
TurboTrain
ax6761
Spitfire
arclight
Mike Ely