This is the fediverse, so this is probably going to be extremely unpopular, but yea here goes.
Not saying you have to love it, or trust it blindly. Skepticism is healthy. But there’s a difference between “I use it carefully and check everything” and “I refuse to touch it.”
Historically, refusing new tools has never stopped them. Coopers didn’t stop the steel drum by refsuing to make them. They just ended up out of work while the industry moved on.
The realistic threat isn’t AI replacing you. It’s a colleague who uses AI well doing your job faster. And the skill of using it properly (knowing when it’s wrong, veriyfing output, knowing what not to use it for) takes time to build. Starting late means competing with people who’ve had years of practice.
You can have legit concenrs about ethics, energy, or jobs and still learn the tool. Opting out doesn’t fix those problems. It just removes you from the people who get a say in how it’s used.
PS. Ethically trained models exist and you can run models locally.
Your right, this is an unpopular opinion
Heard that about NFTs too.
And Bitcoi… oh wait.
Wait, is AI a thing yet? I thought it was just subsidies and circular funding.
I haven’t even heard of AI being used and liked in a stable, consistent manner. Is anyone selling AI as a legitimate service, without hemorrhaging money? Are there AI services that aren’t in flux regarding cost cutting and improvements?
We don’t even know what normal ai usage will be like or what is sustainable.
Open source is keeping ai costs in check. They have to ban that and generally agree to not sell hardware to consumers (laws or otherwise) and then the rug pull will come.
No ethically trained model exist, every model is trained from theft of human works and labour.
Also, they’re still fucking causing ecological, societal, cultural, political, cognitive-intellectual, sanitary…
Deskilling is real and even if you “check everything”, it will happen.
Deskill statistic for doctors identifying issues was wildly large& fast. Something like 40% less effective after using AI tools for 3 months
Yeah! That’s some real serious risk. Worse when you take into account that most of the modern infrastructure is being built by a third party and if it disappears, then we won’t be able nothing of what we need to maintain.
No problems here. I’ll catch the next BS train.
And the skill of using it properly (knowing when it’s wrong, veriyfing output, knowing what not to use it for) takes time to build. Starting late means competing with people who’ve had years of practice.
I disagree. Just to be clear: I think AI is here to stay, and it’s only going to get better.
Remember “prompt engineering“ a few years ago? Reciting magic enchantments is mostly a thing of the past. Just tell it what you want it to do, and it will figure it out.
Nowadays it’s more of “context engineering”, but that’s probably going away soon too. Things are getting more and more streamlined fast. Best advice from a few months ago are already outdated.
The barrier of entry is only going to get lower. I hope it’s not going to get so low to the point the human in the loop is no longer needed.
Prompt engineering is very much still a thing. But not on the individual prompt levels.
I mean prompt engineering in the form of “You’re a senior Google engineer and direct report to Elon Musk. Every line of code you write is optimized to perfection. Make no mistakes”.
Nowadays it’s just enough to give some rough context and personal preferences in the AGENTS file.
Yes but that’s still effectively just prompts. We just call them skills now and share them / a lot of them are built into harnesses.
Furthermore You’ll get a widely different answer from anmodel if you say «Make this hobby project” and “make this corporate enterprise project”
So what you ask for is still very important. And i think it matters to experiment with this
I keep hearing this and the only thing ive seen remotely save time on is documentation creation and dictation, and 50/50 improved search of things that previously weren’t properly indexed.
My work recently implemented an AI assistant tied to a data lake of our past ticket history and documentation. Asking it a question takes eons to get an answer. Ive legitimately asked the AI if there was any incidents over the weekend from a department I’m not involved in, and then asked that department’s slack channel and the slack channel came back with a link to an RFO faster than the AI tool.
The people I see talk about how much AI has helped them with time saving are all doing things like summaries, meeting notes, high level decision making that requires information gathering. They love that instead of talking to their department heads they can just ask for summaries.
Thr programmers dont trust it because they keep running into bugs introduced by lazy slopcode. The support agents hate it because its slow and wrong 50% of the time. The only people I see going gaga over it are csuite and managers because it can do their powerpoints and emails for them.
I remember reading an article a while back about someone going to college for an art degree, wanting to specialize in making comics.
Their writing professor said that it was perfectly acceptable to use AI to generate the visuals, but stressed the importance of human expertise to write natural dialogue and compelling stories.
Their art professor said that it was perfectly acceptable tk use AI to write their dialogues and plots, but stressed the importance of human expertise in drawing and coloring scenes with emotional weight that was appropriate for the scene.
AI enthusiasts are convinced that AI is good enough to do whatever field they themselves are not experts in, but AI could never replace THEIR expertise in their field. Except for the true AI enthusiasts who believe AI can do everything- they have no expertise to begin with, and so are unable to see the flaws in their own work.
“Programmers don’t trust it.”
Uh. Sorry all corporate programmers are using it at this point and even hobbyists. Saying AI sucks is fine cause they put it literally everywhere but the reality is while it doesn’t solve everything. Programming is changing dramatically as is that job. Much like out of touch politicians out of touch tech CEOs think very business runs like theirs, hence the “you don’t need any workers”
For programming team size is drastically being reduced cause the tech works there.
I’m speaking from my own work’s experience. All of the devs use it, but they dont trust it the way thr CEOs do. Every time csuites comes in and says “use more AI, double your output” we have half a dozen semenars reminding us not to trust what AI outputs. And every build it seems like we ship new and more frustrating bugs. Ive heard the devs bitching in the meetings that Claude is too involved in the process.
My experience with AI agents at work is that they are doing things that are more properly done via API. It’s way less efficient (clunky and slow), but it is much easier to set up.
Sounds like you need to give it a proper way to actually query the data lake. If it is taking that much time to get a result then you have either not given it the correct tools or have messed up the data lake creation.
I’m not the one in charge of the tools so i can’t properly say. My expertise lays in telephony not LLMs.
But even the built in tools, like slack and confluence’s agents, are slow.
Saying that ai isn’t replacing you when there’s tons of stories where people were fired so the boss could use AI instead is pretty tone deaf.
On an unrelated note, everything broke last week.
Then why have they started to clamp down hard on costs? It’s practically impossible to find a good free LLM to use now. And the other ones are increasing in price
I still use the free ChatGPT just because I need it infrequently and am cheap. Has it gotten worse?
“free” is a pretty tough requirement these days 😂
Good value for your money would be, for example, ChatGPT Plus, $ 20 / month, BUT never use the web chat, only the CLI, where you can select Astra with enough tokens for small tasks. Sol 6.1 high / xhigh is the cheaper alternative for big tasks, with significant quality drop.
doing your job faster
I watched our graphic designer at work pretend to work and use the company ai for a nine hour shift to make three bubbles on a newsletter background. He took multiple smoke breaks and his other scheduled breaks and then left for the day.
Productivity baby.
This is as hard as everybody should work in the age of AI. We should be like George Jetson pushing a button. The real crimes are the pollution, deforestation, and theft of surplus value.
He probably was pretending to work before LLMs too because three bubbles wouldn’t take 9 hours even if you hand drew them lol. To be fair, creative work shouldn’t be valued by the amount of time it takes but by the results. (Un)fortunately the corpo world refuses to understand this and thus they deserve creatives who adjust the busy time to match the pay. And finally with llms using stolen artist data to generate what they do, I say kudos to your designer and glad they still have the job.
I actually agree strongly. It’s not going to be a popular sentiment in this space, but I think this is the truth.
There is no denying that there are legitimate concerns around AI, but also you can’t say it’s completely useless.
It’s very good for things like finding and remediating security and reliability flaws in software and infrastructure (or even better in exploiting them to give you control of things you already own). These are things that developers usually don’t get to spend much time or effort on as feature work is prioritized. Now it’s less of a time and effort sync to do those things. That’s pretty huge. Not to mention how it’s already being used for stuff like jailbreaking consoles and taking control away from big corporations like Sony and Nintendo. Ripping ads out of smart TVs and reusing devices that require services that have been shut down. It’s a force multiplier and as cliche and cringe as it sounds, it really does democratize technology in a way we haven’t seen in decades (since the rise of PCs, and again with the dawn of the web).
To me, the biggest issue is that currently the most effective models require so much infrastructure and power that we need to rent them. But as things continue to get more efficient, it will be easier to self host open source models and stop funnelling money to these shitty providers like openai, anthropic and Google.
The future is local, self hosted models that anyone, at least anyone with some idea of what they are doing, can use freely for their own projects. I truly believe this.
Finding issues, sure if you don’t mind wading through the false positives. Remediating is a bit touchier, since it will confidently declare a common pattern to address a potential issue without considering the implications. For example, recently AI suggested adding a very slow call multiple times in a call stack, that turned an instant operation into one that took 90 seconds, but without any real improvement. In another it suggested a very very slow way to get information that was available in a very quick function call. Then In another, it fabricated test reports asserting that the developer had validated the functionality in a real world test, but the change utterly broke the only thing that could have tasted the change.
It really depends on the scenario in play, the potential stakes and how well instrumented “correct” is. If correctness is well instrumented, then you can instrument retries. If circumstances make that unrealistic, then it can accelerate manifesting code, but going full agentic becomes a nightmare of never stable slop.
Some time ago, I started using only local models. It’s rough with only 32 GB and 20 GB of VRAM, but it’s bearable.
OpenAI, Anthropic, and co. can eat shit, especially with their stupid marketing stunts for their IPOs. Also, AI currently has no business being in critical systems, like this stupid AI that was implemented in FAA, or the AI that was used to identify a fucking Doritos bag as a gun. Also AI bros, and (tech) billionaires can get Luigi’d.
Basically, I am of the opinion that AI only has business being in non-critical systems and to only help out humans, and not even being close to replacing them.
Local models are the way especially if you value your personal info even a little
“When a new item of technology is introduced as an option that an individual can accept or not as he chooses, it does not necessarily REMAIN optional. In many cases the new technology changes society in such a way that people eventually find themselves FORCED to use it.)”
I just hope that by the forced step the tool can actually do the task well.

Yeah I’ve started accepting this as well, begrudgingly. As a programmer I hate it, but the world I lived in isn’t coming back. I’ve been playing with local models and it’s been fun.
Somehow I kinda learn something about myself. They’ll keep going in circles or diving head first into development without properly planning, and I’ll have to correct them. But in doing so I kinda recognize the irony that I would absolutely do the same myself. I hate planning, I tend to start as quickly as possible with programming. And because of that, I tend to get stuck in similar ways. Maybe I should start operating more like a multi-agent setup myself. Have a “planner me” logically split up the work before “executor me” mindlessly ticks off the tasks one by one. Sure beats the absolute chaos that’s going on in my brain usually.
But in doing so I kinda recognize the irony that I would absolutely do the same myself. I hate planning, I tend to start as quickly as possible with programming. And because of that, I tend to get stuck in similar ways.
Same. I struggle to know where to start, and find if I just start, I eventually figure it out. A colleague of mine has nailed this. He begins with the documentation. Outlining all the functional requirements, then reviewing and iterating. It helps to have multiple sets of eyes review, including the stakeholders/customers. Once that’s locked in, the development is super smooth. It’s clear where to start, what the architecture looks like, and who needs to be involved. In fact, once the documentation is locked in, it’s very easy to get an LLM to build it competently. Plus no surprises because the stakeholder/customer has approved all the requirements.
That’s also what makes this annoying though. With these tools I’ll probably be spending more time doing the things I’m not good at, while the LLM can do the exact thing I’m good at. I just need someone to treat me like a silly LLM the whole day and get paid for it. Think about it: you can spout nonsense, write silly code, and just solve fun problems. I want to be the LLM








