this post was submitted on 21 Sep 2026
455 points (97.5% liked)
Technology
88199 readers
3274 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related news or articles.
- Be excellent to each other!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, this includes using AI responses and summaries. To ask if your bot can be added please contact a mod.
- Check for duplicates before posting, duplicates may be removed
- Accounts 7 days and younger will have their posts automatically removed.
Approved Bots
founded 3 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
I'm finding it hard to reconcile the people who find it useless with those who find it useful. My own experience with languages is that it speeds me up for tasks I have the skills to verify, but with ones I don't it can slow me down doing the verification and fixes.
I know software developers in a Fortune 500 company who swear that overall they are far more productive with "agents". I don't know if it's a highly tailored system rather than the more all-purpose ones the general public have. So I do wonder if there are use cases for it.
It's ruining the planet and our societies, though. And it's distracting us from what really matters. Net negative for sure.
Edit: they're also not very good at their most basic tasks. Researchers have found that when the major LLMs summarise articles, it misrepresents the facts 51% of the time.
It's like a flamethrower, if you're clearing brush it helps you go ten times faster. But if you dont know what you're doing you'll burn the whole property to the ground.
The issue is that the executives only see the potential to speed up, and not the necessity of having skilled people who can do that verifying.
Agents is just the LLM prompting itself with subtasks. This can be helpful to compartmentalize larger and more complex tasks. I find the tool calls most important. An agent that can compile and test its own code will make fewer mistakes (assuming the tests weren't also written by LLMs but even then).
So there's a self-correcting feedback loop that will improve quality, but is also a lot more expensive. The problem is that the temptation is huge to give more and more complex tasks to the agents and then they go off for 30min or longer and may or may not create complete and utter bullshit because something led them astray or they forgot something one agent did wrong and never fixed it. The end result may work, but it may be less efficient or come with a share of dead code that's never called etc.
You end up with hundreds of lines to verify. You could just ask an agent to do it, but at some point adding more agents to the pile is not going to help... and probably takes longer than just reading it yourself.
Yeah strong black/white opinions on AI annoy the hell out of me. I find some AI tooling really helpful like fancy line completions and using conversationalLLMs as a last resort to solving problems, but I hate anything to do with agents or vibe coding in general if only just because I don't want the field I enjoy so much being entirely replaced with being a project manager for a computer.
It makes me frustrated trying to talk about this stuff on Lemmy or other places though because it always ends up being two battling sides of "A SINGLE GENERATED CHARACTER MEANS THE ENTIRE PROJECT IS SLOP!!!" and "ANYONE WHO ISN'T VIBING OUT MILLIONS OF LOC WITH 100 DIFFERENT AGENTS GOING AT ONCE MAY AS WELL JUST KILL THEMSELVES!!!"
I can kinda get the sentiment behind that polarization. There's such a strong drift towards LLM use that it takes hardline stances to try and slow it down
I find it can speed me up, but it's very easy to become lazy and find myself asking the AI to do things I could do perfectly well myself, quickly and using far less electricity. And then before I know it I have a huge pile of slop that I don't know my way around, and I spend the time on the other end trying to understand and check it all. It feels amazing up front, but you pay later.
For corporate software, there's so many bugfixes and features, which require so few additions/changes that it's faster to do it than to write the prompt.
But obviously only if you're able to hold the entire codebase in your skull, but a proficient developer should be able to do so.
If people begin outsourcing these changes to LLMs, then they're cooked and no longer able to verify the output.
Yeah, it seems like a new type of judgement is needed that I don't normally encounter in a day. Like, I use a hammer to hammer nails, then I put it down. I don't try to cook with it. This tool has a weird, abstract nature to it where you can drift between using / misusing / abusing without necessarily noticing. A strange position to be in while trying to work at the same time .
My code adjacent work is much faster, but programming itself is about the same.
So if I need to make a data visualization or UI for data input it’s way better, but if I have to do an analysis or build something it sucks all the knowledge out of the task and takes twice as long. At least. Because unless I’m forming a strong understanding of the information the task can’t progress.
I have no idea if LLMs actually make me faster in the aggregate, but I’m more interested in using them to be better. For example, having a LLM review some code, or share a plan for and have it point out all of the flaws in your reasoning or alternatives you never considered.