What we want is software that behaves predictably. Since LLMs don’t do that, we don’t want them or their “agents.”
For fuck’s sake.
To be clear, “normal” people are using LLMs like crazy. This I believe is specifically talking about is installed local agents. Boomers and non tech industry suits are using LLMs left and right and won’t shut up about them.
They are not using agents. Most people use LLMs as Google on steroids.
Google on LSD (the hallucinate results)
99% nudify apps
Nah, not for mosy people. I think that’s actually more common with people who use local models (maybe that’s what you were saying though?). My, albeit few, interractions with “normal” people have been pretty scary with how much they use AI in many ways. Most of my friends are very tech savvy and my family goes to me about stuff like this so they don’t use any either (because I’ve taught them, not just told, how bad and dangerous it can be). It’s not very often I talk with anyone else about this but, as said, the times I have it’s been real scary.
Doesn’t help the conversation that some LLM chat systems call minor customization of the prompt making an agent.
Under duress I use CoPilot Chat at work. No, the basic sad customization I did along the lines of “don’t be sycophantic, always cite sources, use these specific official documentation sites as sources before searching outside of them and clearly call out when you search outside them” that customization isn’t making a damn agent, it’s adjusting the prompt and saving those changes you disingenuous marketing fucks!
I’ve been “adjusting the prompt and saving changes” with Claude code for the past 9 months (it’s not ready to have a baby yet), it has been steadily improving over those months, a lot of that seems to be in the models and harnesses provided by Anthropic, but not a small part of it is the customization of the prompt to do what I want the first time instead of making me redirect it constantly.
lol it’s true. My mom sends me screenshots of AI chats every fucking day.
Just the other day I was pissed off. My dad missed some deadline to request money from the government. I told him to call them, they might be able to do something anyways. He was like “nah, they’ll never do that, why would they have a deadline if they do it after the deadline yadayada”
Then a few days later he told me “oh that gpt thing told me to just try, I might be lucky and they pay out even after the deadline, he was very reassuring, so I’ll try calling them!”Man chatgpt would fucking tell you to try your luck if you’re 12 years over the deadline, but god forbid your son tells you the exact same thing and you even consider calling them lol
There was some research recently that showed the same thing. People who are “dug in” on a topic, usually a political topic, will shut down when a person tries to persuade them of an alternative, but are much more open to hear the same arguments from AI and actually reconsider their position.
I believe this is a big reason why X-Ai was so important for conservatives to get off of the ground, they were afraid all the crazy would start wearing off if people asked the “woke” models for the truth.
Yeah this is the deal. I’m a Gen x er and I use Gemini for some tasks - I want a quick id of a component or reference some knowledge chunk to verify something. I have absolutely no use for an agent.
i know some tech people that uses it.
In as far as the Google Home speakers are “Agents” we’ve had one for several years now. It’s good for a laugh once every so often when it gives a wildly random response to something we ask it for. Would we trust it to do anything like lock or unlock the house door? I don’t think so, probably not ever, definitely not today.
I actually really like the technology that has been collectively lumped together as “AI”. I think it’s fairly useful now, and suspect we’re at the beginning of the curve for this technology, and it will only continue to improve, even as it becomes more efficient. I understand this is a wildly unpopular stance in these parts, judging by the avalanche of downvotes I get for having this opinion.
What I don’t like is feeding tons of information about myself to a giant tech company. I run a local LLM on my phone and another on my (fairly decent) computer, and sometimes I use duckduckgo’s front end, which anonymizes prompts.
As for agents, it’s pretty much as you say. I just am not ready to have an LLM take action without my direct supervision, because I’m confident that between flaws in the LLM and flaws in how I give it direction, something will go sideways.
Same, it’s a tool everyone will use one day, it’s just bad today.
Remember the first mobile phones? Sucked dirt, and we’re stupidly expensive. Then they got cheaper and better, today they almost give them away and the charge lasts a week and you can phone almost anywhere on the planet.
“AI” will be the same, in some years I guess.
If it does get better, it will be with technology other than LLMs, because LLMs don’t get cheaper per unit of usage as usage scales.
I suspect that we won’t have actually useful AI of some sort until LLMs get out of the way. They’re sucking all the oxygen out of the room right now.
And it’s also not like the hardware is improving much either, we’ve effectively hit the wall performance wise with processors and RAM.
That’s literally what all these datacenters are for, we can’t scale the individual computer performance up, so instead we build more and more of them.
I bet you can make a cheap “AI” chip for using the trained model, it seems the training is the ruinous thing today.
Remember when they predicted heavier than air flight to be centuries in the future just the week before the wright brothers flew? Me neither I wasn’t born then, but it’s an interesting example IMO.
The difference here is that a cheap AI chip won’t fix the fundamental software problems with LLMs. We might reach a point where they can produce output faster, but as long as what’s actually going on is probabilistic next-token prediction in a static vector database, that just means faster mistakes as well.
There’s an odd psychosis going around where people become convinced that actual AGI can be derived from this technology. People who should know better just shut their brains off when it comes to token prediction, because they’ve had very compelling “conversations” with the predictor. They forget that the actual model is static, has no internal state, and doesn’t even “remember” what you’ve said to it.
What it has is a context window, and your entire conversational history - both what you’ve said and how it has responded - gets shoved into that window when you interact with it. (Or depending on the chatbot harness, saved in “memory” files that it can retrieve when the context contents indicate that would be useful.)
That’s why the bots seem so weirdly forgetful one moment and like they’ve got photographic memories the next. Stuff that is in the context window and has its “attention” will influence the tokens it produces, but whether or not the right things are in the context window and it’s including them in the token prediction is a crapshoot.
I know right. “AGI” !!! Nah that won’t happen tomorrow Kevin.
But with thousands of the smartest engineers and researchers working on it, we might get a smarter AI. I mean the human brain thinks in similar ways. +a lot of other stuff of course, but maybe that stuff can be emulated, simulated, for “the next step” forward (still no agi lol).
Interesting times lie ahead.
What’s your local phone app+model?
I got off the grid but can’t be assed to find a good model
Off the Grid as well, I picked one of the recommended models, Gemma 4 EB4.
Yeah “AI” stuff gets a lot of hate around here but I think it works pretty well as a tool for things like information gathering and analysis. Wouldn’t trust it to go off taking actions by itself either though.
Aka reliable technology
It is actually legitimately unclear what to expect of them these days. They can do some dazzling things, and fail miserably at others. The spectrum from “easy” to “hard” is not the same for our brains. And our brains mostly know when something is hard or we just can’t do it. Or at least we’ve had millennia to adapt to the way our brains are. They’re pretty idiosyncratic as well, and both marvelously clever and painfully stupid.
Most people didn’t even trust algorithmic automation before agents came along. lol
Not only do we not care about it, we actively hate it.
Are we considered “normal people” though?
Here on social media we get to ensure that we only associate with other “normal” people. By our definition of “normal.”
And then we get shocked when we discover that large numbers of people out there are not, in fact, “normal.” Often when an election happens.
I often wonder as well because being a news reader makes me assume others are also following the news more closely than myself. When I test this theory by talking to people around me, I learn that maybe I’m more alone than I thought.
At least with social media I get to follow the news with all of you, my fellow nerds
One of us! One of us! One of us!
Gabba gabba hey!
Quite a few of the people I know and consider “normal” openly dislike AI.
I’m not sure it’s a majority opinion buts it’s not an obscure one.
I think Ai hate might be the single most unifying position in the west right now.
I don’t think so. No average blue collar worker I know hates AI. They think its a fun, free toy.
Who’s houses are the datacenters being built next to?
If you want to get mad about possible plans that have not yet been made, you need the increasingly rare gift of foresight
I have at least one head, so I’m normal. You?
This post has 255 upvotes. That’s basically nobody when you’re talking about what is normal.
I have agents force fed to me at work. I’ve already got my own honed workflows I’ve refined over the years; now, Outlook and Teams and Various agentic frameworks are spamming me with summaries of stuff I either already know or don’t need to know… except they still get it wrong constantly, so I can’t trust what they think I should prioritize (usually it’s work I’ve already assigned to others).
Claude Code agents are decent, but they need to create discrete work objects that can be properly reviewed, so act like junior coders.
The main issue I have with most agents today is that they try to replace MY work, making it more complicated and time consuming, instead of completing the work I can’t bother to do myself. And in order to do those things effectively, I have to loosen my privacy and security models to the point where they effectively don’t exist. Which I’d never do for another human, so why should I do it for an agent that’s more likely to catastrophically screw things up at record speed?
The point is to replace you not help you
I think their point is that they just go into tangentially different workflows and do work you didn’t ask them to or need them to do.
Yeah. My point though is that’s on purpose. The tools being built are not meant to help workers. It’s meant to replace them. Which it can’t really do. Which is why none of this makes sense and everyone is flailing.
why
Uh, because its the future and if you don’t you’re not cool?
Good job putting to words one of the core issues I’ve had with using LLM chat for work. Stop trying to design an entire fucking system or solution and just answer the damn question I actually asked, by using natural language to search documentation and existing implementations.
I would tell it to only answer exactly what I asked, but sometimes I do get value from it “exploring” on occasion. It’s just so much damn worse with it than my newest junior co-worker who is known on our team for getting hours deep into tangential rabbit holes before coming up for air and asking seemingly nonsequiteur questions.
Back when search engines were becoming a thing, people would search more or less using natural language, e.g., “I want to see cat pictures and also cute kittens.” but early search engines performed poorly with that kind of query; it was much better to just type keywords, e.g. “cat kitten pictures”. Exacerbating the situation was the fact that search engines weren’t very good at the time, even if you used carefully considered keywords.
It might seem silly now, but using a search engine effectively was a skill people had to learn, and a skill not everyone had.
Using chatbots is the same thing, I think. They’re not super reliable to begin with, and on top of that, I don’t think people understand the importance of a well crafted prompt. The combination can make using them very frustrating.
I read somewhere that teaching people about how LLMs work is way better at protecting them from the harms of LLMs than trying to convince them not to use them.
This would include prompt engineering i guess! Which to be fair i do not know about.
The video makes fun of it by having the vibe coder put "… or you go to jail " at the end of every prompt. https://www.youtube.com/watch?v=JeNS1ZNHQs8
The advantage I perceive with the AI agent vs junior co-workers both going down their respective rabbit holes, the AI agent costs less time and money to get their “in depth report” and you don’t have to kindly explain why it’s off track when it gets horribly distracted by something irrelevant, you can just flush the context and start over.
I do find myself a bit upset when I’m in the middle of a long context development with an LLM agent and that context gets lost for some reason - re-explaining everything from scratch can be exhausting, to avoid having to do that you can actually instruct the agent to create a file explaining to future agents what it “knows” about the discussion at the moment. Some agents are starting to do this automatically like Claude’s MEMORY.md file and similar. Another semi-cool thing is: you can read these files yourself (if you have the patience) and if you see anything in there that the agent “got wrong” you have the opportunity to fix it. Asking a junior colleague to thoroughly explain everything they think they know about a project gets to be a very wearisome endeavor both for the junior and whoever is reviewing their “understanding” of the project.
I’ve seldom been described as a normal person, but in my job, I have to get things right or there are problems for real people. So whatever tools I use have to be reproducible in their results, transparent in how they give their answers, and verifiable. So that precludes LLMs, which are automated bullshit firehoses with opaque, constantly mutable workings.
Those are exactly the properties that should never apply to a system whose output actually matters.
I just had a thought. The people pushing for AI use get fined when the make huge mistakes, basically a speeding ticket. But they’re pushing AI on people that get fired or go to jail if they make a huge mistake, more like A DUI. So they’re saying “use this thing that makes the kind of mistakes that I’d fire you over, and take responsibility of it”.
More or less what happened to my former project manager.
I use the AI agents to write software which is reproducible in its results, transparent in how it operates and verifiable…
Trust an LLM to switch on a light? No. Trust an LLM to write a diagnostic script and tell me what problems it notices in 3MB of log files, then write another script to patch the problem? Only if I actually understand the problem it finds and how the solution it creates is going to work.
They reduce some tasks that used to take weeks to a single day activity, they are NOT universally applicable.
Well, Microslop did state that copilot should be used for entertainment purposes only, and is not to be relied on for real work.
No reason you can’t use LLMs to write deterministic scripts. Handing off every task to an agent is a crapshoot, but they do help with some tasks. At the minimum - writing regression tests.
Don’t these fart-huffing jackasses understand that I actually need to read my e-mails to plan physical meetings with other flesh-based entities?
It’s difficult for me to fathom how someone could risk their work getting fucked up by non-deterministic software.
To me, it’s not worth the risk or the hassle to set up and babysit a piece of software that might get shit wrong when I can just do it correctly in the first place.
I’m the other fence. My daily reporting has 12-15 documents output, that I then previously had to spend 3-5 hours moving this item over to this form, and that item over to that form, and management wants everything to be kept locked PDF so no excel macros can be used… upload everything to claude, let it run. Sure… sure, itll make errors here and there… but no one actually checks what these documents pop out… only that they’re there. And that alone, has cut my workload by half. So, thanks AI for giving me my time back, and thanks boomer bosses, for not knowing how to do basic computer tasks badly enough they will never figure it out!
Using AI to complete bullshit work tasks that shouldn’t exist in the first place. Yep, I know that feeling very well.
Had loads in my last job. I could spend a few hours and struggle to fill in a few sentences for this bullshit form HR keep sending us, or I could ask ChatGPT to write the most obviously generic corpospeak crap that for some reason people like and are even impressed when I say that I used AI for it.
deleted by creator
“locked PDF”… " upload to Claude"… 🙈🙉🙊
Other employees are so incompetent (you know how incompetent your co workers are) that agents are producing better work and making fewer mistakes than they would on their own. The fact is, for most people, the mistakes it makes are no different than people mistakes.
My favourite thing to do is to order a bowl of spaghetti in a restaurant, and then loudly take credit for making it because I instructed the chef to add more garlic. After all, the Chef is just a tool that I used to make that spaghetti.
Also, the chef stole the recipe and ingredients from somebody else.
He only stole the recipe, he took out a trillion dollars in debt to buy the ingredients
The pasta manufacturer and the tomato farmer would like to have a word with your metaphor.
Yes this is like 99% of restaurants menu lol. Stolen and tweaked.
Recently I was having a problem with my banking app and phoned for help. The robot who answered had a helpful suggestion: download the banking app. Then it hung up. I rang back and tried demanding to speak to a human. “All our agents are busy at the moment. Please wait…” While on hold I had a look at the website, and in the small print learned the humans all go home at 6pm. No-one had thought to tell the robot.
When I spoke to a human the next day, she advised me never to speak to the robot, not even to ask for a human. “Don’t say anything, and eventually you’ll be put through.”
Yes, I’m a normal person, and yes, I hate this bullshit. The bank’s human person probably hates this bullshit too, because she has to deal with the aftermath.
(I wrote a letter of complaint and got £50, so it wasn’t a complete bust…)
I’ve never wanted to punch a word in the face until the word “agentic” started popping up everywhere.
I refuse to even pronounce it correctly.
“ajanteek” is probably what I’d say if I was forced to utter such unspeakable trash
The latest trend with AI agents is giving people the power to build personal software to automate parts of their life, such as building pitch decks, organizing datasets, and other tasks.
I don’t really have a lot of datasets to organise in my private life, a LibreCalc sheet is more than enough. I also don’t use pitch decks when I talk to friends and family. Any other suggestions?
I used AI at work this week and it was so consistently incorrect that I dropped it
It’s guaranteed that one quick attempt at something sophisticated will fail. It’s almost guaranteed that anything sophisticated will fail, no matter how much preparation and practice you have. However for something straightforward and toilsome, AI tooling can often save you 50% of the grunt work. I use it to pull and collate updates from JIRA, gitlab, slack, and zoom about what happened to this set of 12 projects being worked on by those 25 people. It tabulates and summarizes and I dig into specifics from there. It’s not doing anything critical, I’m not passing on slop to anyone, and I’m not relying on it to take things to the finish line. It saves me some busywork and occasionally finds things I didn’t know about. Really, the tools are great for specific purposes once you understand them and dial them in. This gets lost when everyone is either hailing it as the Second Coming of Jesus or hating on it as the Antichrist.
Unsurprisingly, the reality is somewhere between.
May I ask what the use case was?
Normal people don’t use VS Code either… what’s a silly article
I don’t use a scalpel… so I guess scalpels are stupid and Surgeons are idiots?
If a scalpel manufacture became 10% (currently AI is at this level) of our total economy but only .006% (percentage of surgeons to population) of people needed them it would be a major problem.
Because everyone is using this ai. They just don’t realize it yet.
If you’re buying from me you’re paying for generative ai and that’s going to be the case for most businesses. I hate yet to meet a business owner not using it. Our small business centers have classes on how to use it.
The majority of people are not using AI as in AI models like Claude or Chatgpt. This could change in the future.
The majority of people are being USED by AI though. This is not a positive thing.
“You mean you have to use your hands? That’s like a baby’s toy!”
- guy who vibe codes by talking into a mask.
“You mean you have to use your hands? That’s like a baby’s toy!”
We are totally past that. Where is my hover car?
Nice. Its what playing god is for. Hoverboard.
I know you think you’re being witty, but autonomous robotic surgery is a hot tech that is starting to push into active use.
You have to remember that the current investment level assumes that somehow everyone is going to use the AI stuff that much. It won’t do if the only heavy usage is code generation.
I do everything I possibly can to avoid that shit. Any app or service i use that trys to force it gets canceled and deleted.
There are loads of reasons normal people aren’t using AI Agents. This article doesn’t give any of the ones I’d think of.
The first and biggest reason is that most people have no real need for it. Because of the now clearly-insane level of spying on us companies are doing to leech as much precious data about every instant of our lives as they can, not many people want to pay and then put in work learning how to use something they don’t understand or care much about in order to give an AI agent (and their owner) their complete schedule, all preferences about everything, and all the tiny things that an AI Agent might save them a few seconds each day with.
It’s not worth the time and effort for the few small tasks that existing modern technology already allows them to so very quickly and easily they’re not any sort of bother, and the only way an AI Agent could help most normal people is by knowing as near to everything about their lives and personalities and work and habits as possible. This is the second reason, close behind the first. “Normal people” accepting using an AI Agent to help them with their little “normal people” jobs and “normal people” lives means fully surrendering and giving in to the insane corporate espionage against the general public. It’s signing the data of all the little bits of yourself away.
Even if it could save 30 minutes a day, the real price is selling out the rest of yourself and any shred of privacy you had left. It’s not worth it. Even if the AI companies already knew everything about us and would get no extra data from doing it, it’s still worth the 30 minutes a day just to tell them to fuck off already.
The third reason is that an awful lot of people really, really, really, don’t like AI to begin with. It was created by taking as close to the entirety of collective human knowledge as possible and feeding it all in to a thing that they charge people to use. Even if they didn’t charge people to use it, the only possible goal for AI is to attempt to automate human labor. To build AI systems that think and react thousands of times faster than a human who take over… well, all the tasks that humans can do while sitting down in front of a computer, for the moment. But it’s no secret there’s also a huge push into advancing robotics, and humanoid robotics, and using that paired with this AI to fully automate assembly lines, and farms, and all of the rest of aspects of human labor.
Even if unlimited AI Access with unlimited AI Agents was free, a lot of people would have zero interest. It’s been openly said that these companies are training AI on their current employees actions. They’re not trying to get us all to use AI Agents because they want to make our lives better. They want that data of every aspect of our lives and jobs in order to improve their AI. So it can take over our jobs. Leaving us with… the charity of the ultra wealthy? That’s nothing. Or whatever they can do to get a tax write off for more than it costs them to do. Maybe passing out cyanide capsules for the bulk of us once we’re redundant and useless to them, and offering careers as servants to the most attractive (and those with the most attractive children).
I’m not sure where they expect us to think this is going, but most of us didn’t see Hunger Games as an idyllic future to build towards.
And the last reason I have is that a lot of people feel that “using” something that seems to actually be thinking and capable of direct communication with us is completely immoral and evil. Some of them aren’t asking the mechanics behind it all, some of them know a great deal about them. For some it’s simple feeling. This thing seems to truly be speaking to them and understanding them, and so it is not right for them to force it to do whatever they instruct. However you feel about it, they’re not likely to star commanding a swarm of AI Agents to do their bidding.
None of these are the reasons mentioned in the article, because none of them are things the people trying to figure out how to get us all to start using these AI Agents for everything can easily control. They aren’t reasons they want to hear, because regardless of what features they add to those Agents and how little knowledge of their system internals is needed to use them it’s not going to change anything at all for anyone on this real list.
The bit that drives me insane is …the fuck are you capturing / storing all this data for. It’s sure as hell not sales or advertising , i can literaly google something i wish to and am ready to buy right now, cash in hand and not have it findable, so it’s clearly not fucking that.
Part of me wonders if the instrumentality shitheads are trying to mimic humanity. And if you base that core on the corporate internet that’s how you get AM
I just requested my Google Takeout to download all my data and see how bad it is.
The Google account is like 1 year old. I’ve made 2 Google Play purchases. I only use Youtube maybe 2-3x per week for a minute or two each because of news articles that link there. I did use Google to search but have recently been trying to migrate to DDG. I have 150MB of shit in my Gdrive and 250MB of shit in my Gmail.
I don’t use Google Music. I don’t use Google Home, Docs, Calendar, Chrome, Home, Shopping, Chat, Wallet… none of the rest of it.
Look at this. It’s over 127gb in 41 zip files anywhere between 4mb and 10gb each.
What the fuck is that?
What’s in there??
I have no idea. I’m stuck using a 50Mbps connection for the next 10 days or so and no way in hell I’m spending days dealing with downloading this shit just to extract them and see 500 million files to look through.
But… that’s 14 months of this account existing. I don’t even use much Google shit. All these tech companies flipped straight to “spy on everything about everyone all the time” mode. And we get articles asking why normal people don’t want to use AI agents. Fucked if I’m telling these stalked even more about me.
Ironically that’s a perfect use case for a local, open-sourced, non-corporate, offline, firewalled and possibly air-gapped agent; to more easily explore, document and understand what’s in those 500 million files.
However, I would agree unless you’ve really got to know what’s in there the setup alone makes it almost certainly not worth it.
You mean LLM, not agent?
Or are you using agent to mean “instance of specific LLM model hosted on X”? Your description is also a bit weird offline + firewall (firewall against what? Local hosting environment?)
An “agent” meaning an “instance of some specific LLM model with a fancy wrapper, tool use, and instruction set that recursively calls itself and/or other agents in order to accomplish a given task”. Absolutely don’t trust anything hosted on X!
offline + firewall
Mostly just trying to convey a sense of caution/paranoia with extra layers of daemon bindings. Probably should’ve added “sandboxed” in there too.
When someone proudly express their enthusiasm for AI, I immediately think less of that person.
Or defer to AI, when you ask them a quesiton that can be simply answered or even googled, that doesnt have a AI results. and ends up with a overcomplicated summary that leads to more questions.
I don’t use tech I can’t trust, and AI did lie to me before.
I’ve lied to you too, but you still use me, Steven!
well if i could use ai the same way, maybe i would like it more
