- cross-posted to:
- [email protected]
- cross-posted to:
- [email protected]
You must log in or # to comment.
I’m going to lose my mind if I get any more support tickets where someone has cut and pasted the output from Claude instead of telling me what the problem actually was.
My job has basically turned into guessing what bad info Claude has given
Just complete the AI slop trough and copy and paste the ticket back into Claude to fix and open an MR and then have Claude review it and merge and then when there is a bug repeat the process.
I honestly don’t know what people are thinking, both employees and employers when people start copy pasting LLM shit. Why do I need you if all you are is a glorified macro? I strongly suspect the human is there to be the fall guy when the LLM invariably does something really stupid.
Tom still used AI — and that’s fine. Encouraged, even … This is kinda only half serious. AI is great — use it all day. Just don’t be a pipe. 🥩
That site is still promoting AI. So close, so damn close
It’s of course written by someone that didn’t read it themselves… sigh
I strongly suspect the human is there to be the fall guy when the LLM invariably does something really stupid.
This has a term, actually! “Accountability sink”.
Yes, like a heatsink in a computer, your job is to soak up all the damaging and unwanted aspects of what the “real” and “valuable” main components decide to do, so that consequences never reach any decision-makers.
Fun!
The fabled meat proxies.
I have narrowed the containment of the isolation.
Now I’m verifying the resolution path of the indicators.
Permission to accumulate the resolution of the transformation.I know what all of those words mean individually…
Work slop! Just love people using LLMs to hand the heavy lifting over to me.
My god, had someone send in a merge request with utter garbage, with Claude describing some nonsensical “root cause” as to why a function didn’t work. Claude made two changes:
- If the current scheme failed, use an entirely different scheme it claimed the software supports that does not work and was never documented as a way to work.
- If that scheme fails, swallow the error and don’t report the problem, leaving the problem silently in place but out of the user’s view at least for a short while. This is actually the change that ‘fixed’ the problem.
So I needed to know the debug data for what really went wrong, but the submitter got all pissy and said Claude already sorted it out and why can’t I just accept their slop merge request as-is, refusing to believe that all it did was catch and silently drop any and all errors associated with the code instead of fixing it. Wouldn’t run the procedure I stated would show them they had messed up data as a result of their ‘fix’.
Looks like there is an angry cryptobro downvoting responses who dare talk shit about claude in this thread lol
I like to tell them, “well my Claude said this.”
You can just point out the issues to Claude and get it to spew out AI slop on how those issues are the worst thing in the world, horrible bad bugs.
Fighting fire with fire. Except you’re actually trying, which means you can’t really loose.
In their defence, the one thing AI does do well is Code review. But, it makes too many absolutely sh** assumptions when helping customers troubleshooting.
I feel the problem is that because its a paid tool, it’s not a good look if they simply acknowledge “we’re guessing, and its a crap guess”
I don’t even get tickets. I get tagged into a slack thread full of slop trying to guess why I am here and what I’m supposed to do.
That sounds awful, my condolences
Ah, actually, caffeine can be really contra-productive for autistic people.
I was only ever observed to have autism, but I highly suspect that I’m audhd. What kind of hell is this where caffeine can help my ADHD focus but harms my autistic focus
Caffeine be like:
I’m doing thousands of calculations per second and they’re all wrong
If it’s any consolation, I’m AuDHD myself (ADHD diagnosed first) and I would have never made it this far without copius amounts of caffeine.
Edit: To expand on this, my autistic tastebuds have rejected the taste of coffee until I was like 21. Imagine the astonishment following my first cup when I was bustling around actually getting things done for once.
Vibe coders not understanding their code because they didn’t write it.
Me not understanding my code because I wrote it a month ago.
Replace/append with medicated ADHD people
a fraction is being too kind to us
from what I’ve heard from my software engineer friends, a decent coder with standard knowledge of CS and the language they’re using will always write better (more optimized, more debuggable, etc.) code than currently available LLMs
even if the “fraction” is closer to 1 than 0, the heavy lifting here is done by “look what they need.” vibe coders need gigantic climate destroying machines to write code that isn’t as good as what an educated programmer writes running on a burrito and coffee
Depends on your domain. Devops is mostly a solved problem terraform is well documented and most infra follows a far more simple pattern. It’s not replacing engineers but if you aren’t a moron it can be a 5x type tool. It is extremely good diagnosing issues, parsing logs, and frankly writing well formed TF. It’s garbage at architecture and every once in a while it’ll absolutely send you a wtf were you thinking type pr. But it’s 80% good with a 20x in raw velocity.
Our team has fixed up a lot of shitty technical debt, increased automation and our deterministic scripts and workflows, and increased usable test coverage on $30/mo copilot plans.
Unfortunately we live in day and age when corporations prioritize quantity over quality.
And that age is going to be very short as it will end with flaming hacks and downtime.
They always did this. Now they are just doing it faster.
I’m tempted to try to calculate how much energy has been used to produce the currently and previously available library of anime popular amongst coders, because that’s a necessary resource, along with various energy drinks and snackies
I used to write a ton of code while caffeinated.
Now I write less, better code since quitting caffeine.
There was a brief period where I was incapable of doing anything at all between the state of high caffeination and no caffeination.
Be me AudADHD locking the fuck in doing 10 hours of work in 2 hours when the legal meth kicks in. But it’s at 11:37pm and sometimes it’s wikiholes on GBU stabilizer fin evolutions instead of work.
And then I waste the rest of the next 2 days.
What about us caffeinated autistic vibe coders? Slop production through the roof bb!
I put together a new modpack for minecraft this year and from that i can verify that this comment definitely checks out.
Link or it didn’t happen.
Its not a public pack sorry But the changelogs for some mods speak for themselves no human does 20 seperate release updates in a day and comments every sentence with an emoticon
Just caffeine? I thought a sizeable portion of yinz are on adderrrall.
Some of us have to rawdog it, due to the meth myth.
What’s the meth myth
That ADHD meds are just fancy meth, so we have to rigorously select who gets them.
It is just speed though.
Dexy’s Midnight Runners? Adderall is just Dexedrine
Lisdexamfetamine+caffeine, let’s gooooooo
can’t eat, coding…
Hands shaking but mind is going brrrr
It is by caffeine alone I set my mind in motion. It is by the beans of Java the thoughts acquire speed, The hands acquire shaking, The shaking becomes a warning. It is by caffeine alone I set my mind in motion.
Everything brrr if you coffee enough
Even though i hate ai and what it has done to my profession i can confidently say that you are in denial and this is pure cope. Agents can already write much more organized code than any developer in my company. “Knowledge” and “architecture” won’t save anyone since llms already have plenty of that. We became a meat proxy and it spreads in other professions too. What a shame.
You know I keep seeing this, and I keep trying AI, and my experience really just does not track with the grandiose claims y’all AI fans (with 2 comment accounts that have been dormant for six months…) keep making. Maybe you’re some canned code shop with crap devs, but even then I’m real curious what the use case is where AI agents can take over from real devs like you’re claiming. Sounds like marketing hype more than anything.
You may try to attack my character by implying a bot or throwaway account but my point stands and all metrics agree with what i said. I literally used the newest model of a certain company (not naming them) and the agent finished a weeks work in 2 sessions. It even found lib incompatibilities and a weird bug in codebase without even prompting for that. The models became better but the harness got really good now. Your experience may be a skill issue, your prompts may suck or even your codebase might be too big, or cope, either way we are screwed, please don’t give a false hope by moving the goalpost each time a llm evolves further, it’s better for programmers to swim with the wave and evolve with the profession than be stuck in denial and becone obsolete.
Y’all have been saying “the problem is you’re not using the newest model/harness” to any criticism for 3 years now lol
“But have you considered that it might not be a fundamental misrepresentation of the capabilities of our system that’s been borne out repeatedly in real-world testing, but instead a skill issue?”
GPT 3.0 couldn’t create gta 6 using a single prompt therefore all future models are bad. /s Maybe get out of your cave and test it yourself instead of coping in denial.
I have tinkered with it, and I’ve also had to deal with coworkers using it. I have actual work to do instead of playing with toys that make my job harder.
Dude, come on - what are you actually trying to accomplish here? All you’re doing is throwing a tantrum because people don’t agree with you, and that’s pretty unpleasant behavior.
I implied you were neither of those things - I just quite openly implied you’re a shill. Perhaps you should ask an AI for tips on improving your reading comprehension?
Please give me 1 thing that a mid developer can create but the latest model agents can’t. There is coping and there’s delusion, i think you’re heading towards the second.
It’s genuinely depressing you’re so creatively stifled that you haven’t encountered a task that meets those conditions yourself and instead have to retreat to these middleschool-tier insults, but hey, if you’re gonna throw down a mitten in challenge, who am I to refuse?
Novel reporting across diverse data systems! That’s not even a mid developer task, that’s intern shit - and good luck getting an agent to deal with that one lmao.
The more i use it the clearer it becomes it can do much more than a human. One of my hobbies are electronics and embedded development. The llm designed the circuit and even the firmware in 5-6 prompts, that took me a week on a similar project. With clean requirements and scope it may do almost any task, it may struggle on huge projects and monoliths but looking back to each improved model i can guess that won’t take a long time to accomplish either. In less than 2 years we went from buggy gpt 3 to multi agents workflow that work almost autonomously.
That you’ve gone from “I bet you can’t name even one thing AI can’t do”, to being told something it can’t do, and now immediately you’re ignoring that interaction in favor of telling anecdotes about how great AI is? That says a great deal about your conviction to your own arguments.
You basically did tho lol
Hmm. No, but thank you for your contribution!
It’s not up for debate. I’m literally telling you how people are perceiving your response. It’s not something you can disagree with. Just because you can’t understand that other people are reading you differently doesn’t allow you to give it a “no thank you” hahaha
No, (but) thank you (for taking the time to clarify)!
I think the truth is somewhere in the middle, but I could be wrong. I work in a non-programming technical field and I am also impressed and concerned by the progress the models have made over the last few years, especially in terms of being able to “do my job”.
Like the other guy was saying, there are still issues with its implementations, and the second you begin to think it’s perfect or immutable, you notice issues that would have been show stoppers if they made it through. And at the end of the day, at least for what I do, if nobody is here to press the buttons, the AI will never be able to do its thing to begin with.
I also personally predict that the cost of tokens and utilizing AI services will eventually become overly expensive and encumbered, as does everything under capitalism. Name one thing that hasn’t gotten shittier and more expensive overall. The AI companies haven’t even begun to enter their enshittification stage yet.
I maintain that AI in its current form will NOT be able to produce truly novel thought or original work, as it can only put together the Legos that it already has ingested. Perhaps grunt work programming will require less manpower, but cutting-edge developments are still beyond the LLM’s grasp at the moment.
Anyways, I welcome other opinions
Software development is by far the most affected profession. Every time a new model gets released another part of my occupation can be done by it. People are moving the goalpost to focus on every mistake the llm makes to cope and fool themselves they are still not expendable and better than the llms. With enough confidence i can say that the vast majority of programmers are not.
Agents can already write much more organized code than any developer in my company.
Then you must have nothing but absolute shit developers.
I could either take several tedious hours to write out several thousand regex patterns by hand to automatically recognize differing formats of notation for this thing I’m doing IRL…
or I could have an LLM do it in a few seconds, and it’ll get 99.99% of the cases, missing maybe one case every hundred thousand…
take your guess as to which option I chose.
LLMs are a tool. I use it as such.
or I could have an LLM do it in a few seconds, and it’ll get 99.99% of the cases, missing maybe one case every hundred thousand…
Even if the LLM does things perfectly 100% of the time, how do you expect to maintain your skills if you offload all of your work onto a chatbot? What happens when you start offloading more and more until you’re not actually doing anything except typing in prompts?
My skillset is geared towards meatsacks and things that go boom. LLMs have been a massive boon to the 1% of my job that needs a code monkey.
As long as people still talk to each other, and things need exploding, I’ll have a job. Heck, if we somehow get to a point where money isnt needed, I’d still do what I do simply 'cuz I like it.
Isn’t this a case where regex may be the worse choice?
write out several thousand regex patterns
Oh my! This thing is more likely to gain sentience than any llm.
Can’t be worse than when I was an intern, the product was literally based around parsing arbitrary HTML with regex and not even decent regex, POSIX regex from like the '80s for reasons known only unto the programming gods.
sweet merciful torvalds, that must have sucked.
Glad you made it out of that hellhole.
Cheers, me too! Was educational as an intern to say the least. I think the firm went belly-up a few years later.
When I wrote my numerical parser, people told me to use regexes to detect what kind of numerical value is being parsed. I just did it in one go. Only regex-like checking is for ISO datetime values.
Lots of folks here create absurdly high expectations that most coders don’t meet. They are out there but they ain’t average. Getting it 80% correct with good human reviews to catch the trash and using it to figure out what’s breaking quickly vs tracking down logs has been great for our team. I laugh at some of the garbage it spits out and the dumb way you can see shit like, ‘I have to be honest I did zero testing’.
