In addition to making people stupid, I wonder what affect will LLMs like Claude will have on programmers? How will new programmers learn if companies start using Claude?
@grok this true?
Hi Lisa.
Hi Super Nintendo Chalmers.
I’m lernding
I have no doubt “AI” companies are sitting on studies proving their shit causes irreversible brain damage, much like tobacco cartels used to sit on studies proving their shit caused cancer.
By the time the bubble pops and their shit gets properly regulated it’ll have crippled a whole generation (on top of all the other damage like destroying the Internet, causing unfathomable damage to science, culture, and society in general, and infecting any information produced after this shit became commonly used).
I have very little hope for our civilization being able to survive this self inflicted disaster (and given how we’ve squandered natural resources and caused a runaway greenhouse effect that’ll make our world mostly uninhabitable for humans without massive industrial effort that will be impossible after our fall, no new civilization will be rising after this dark age). But hey, at least some sociopath CEOs will have made a lot of money out of it. Who cares if they murdered the future for their short term profit.
By design.
Has anyone found an effective way to pair-up and “learn” the syntax faster/better compared to not using AI?
I’ve written a lot of code in the past, but recently started doing more with golang… and have been using AI for an assist, but at the end of the day (and enough reiterations) - it creates readable and maintainable code. But (unfortunately), I don’t think I could rewrite it.
I was contemplating seeing how I could change my workflow, so I’d write the code, but AI would offer fast guidance.
invest more time before you ask for help. Same as when you’re working with a person.
Also, add instructions to your agent to use a Socratic question teaching mode.
You need to force yourself to think, else you won’t learn.
I wonder… are Google and Bing search indexes being intentionally left to moulder specifically to drive people to Gemini and ChatGPT?
s/could\ be/are/
They’re lobotomising tools. Vibe coding is just shoving one end of an ice pick up your nose, setting the other on a keyboard, and replacing its handle with a mains powered personal massager.
Tech in general is making all of us more stupider, or at the very least, suckers and rubes.
Absolutely shocking.
I wish that I didn’t have to use them, but for basic Linux troubleshooting, it’s easier to ask the chatbot and have it explain itself and cross-reference than to use a search engine or forum. The internet has become so shitty for self-teaching. You either end up on YouTube watching videos that aren’t relevant, your search engine pumps you to garbage on top of ads, or you find a forum from 5 years ago with someone asking the same question, but it was closed because it’s the same question that was asked 20 other times, and all the solutions don’t actually work.
I know that I am robbing myself of this collection of secondary skills, and that part of it is a lack of patience, but the pool of knowledge that is the internet has been poisoned. The only way I even get to useful guides anymore is if Claude links me to them. It’s becoming impossible to use the internet otherwise. Even spell checkers have gotten shittier.
Even in Word, it will try to autocorrect or just tell you something is wrong without auto-correcting, as if I want to search for the word that it obviously marked in red. Even if it offers a correction, it will be one word, and it will be the wrong word. My phone too, constantly changing my words or somehow hitting the wrong letter when it didn’t before.
I just don’t even know anymore. The mental energy to “do it myself” is exhausting. Then, of course, whatever I do or learn to do will be undone with the next update that gets pushed out and changes all my settings. My fucking phone settings have completely changed between when I got it and today. Every app has tied its permissions for functionality to its permission to send notifications, and those notifications are just ads. If I want to order DoorDash and know when it arrives, I have to agree to have DoorDash bug the shit out of me to order food when I don’t want to.
I’m just tired, and if using an AI helps me figure out how to replace my bootloader when I accidentally deleted it, then so be it. I can’t fight every battle.
it was closed because it’s the same question that was asked 20 other times, and all the solutions don’t actually work.
This has absolutely been my experience as well, yeah. I rarely go straight to LLMs first but sometimes I’m not comfortable with simply posting my problem on Reddit and waiting (or posting on Lemmy and waiting forever lol), since it isn’t always that simple or convenient to first anonymize data before getting public help, nor do people even always reply, leaving me back to square one with time wasted. I am by far the techiest person in my company (and I barely know anything relative to full-time programmers), so there sure isn’t anyone at work who I could ask for help, haha.
I don’t go to LLMs often, and they sure messed up a lot even just 2 years ago, but they’ve come an impressively long way from the unintelligible-text and six-fingers era (despite still occasionally messing up nowadays) and have legitimately optimized my formulae dramatically in both function and legibility, such as by showing me
LET()that I didn’t even know existed. While I’ve been meaning to take an online course on spreadsheets and just haven’t gotten around to it yet, I try to learn from how the model did XYZ right away but it’s nice to have a personal, on-demand tutor, as long as you’re always aware of the fact that what it says could be garbage, despite it speaking so confidently. I think for actual, entire vibe-coded products, it’s much harder to learn what the heck it’s doing versus just simple one-liners. Anyway, on-device and FOSS LLMs make for better compromises, I think, though their quality is noticeably poorer.The reason I try to avoid “AI” use is not The Dumbening™ but because of the ecological impact, yet I contend that spending hours and hours trying to traditionally solve a problem in frustration is also harmful in its own ways… It’s a decent tool of last resort and the Stupids are merely the ones treating it as a line of first defense against everything.
The hilarious part about this is that nobody seems to love using these things more than the C suite. Let’s see how dumb they can get before their brains can’t even handle their bodily functions.
deleted by creator
Substituting your own information syntheses, memory and deduction Willa trophy this faculties, leading to people who can’t think. Everything their mind needs to complete thought, particularly complex thoughts, connect to an external device. Their thoughts will resemble Swiss cheese, partial ideas with numerous large gaps.
They can also make you smarter if you use them right. Key is to use local models and not giving the techbros any money.
That’s on my to-do list. I’m currently reworking my entire build because I realized I had enough last generation parts to build a media server. Once I have windows set up to only run on VM and get my stuff moved and backed up I’m going to install an LLM
I recommend Qwen3.6, either the 27B dense or the 35B MoE model. Both outstanding for local models.
What hardware are you using?
I am using qwen3.5 9b. And it is barely working.
I have a Radeon RX 7800 XT.
Qwen 3.5-9b is blazingly fast on it. However while it’s its impressive for its size, it has its limitations. Complex tasks with several steps are too much for it.
So now I run the 3.6-35B model with llama.cpp It’s too big for my VRAM so I had to split it: everything that doesn’t fit on the graphics’s card runs in the normal RAM. That slows everything down, but with the right flags I get a bit over 20 tokens/s.
If you have problems with speed and you’re using ollama I would replace it with something faster like llama.cpp.
Keep telling yourself that
Sure! How much experience do you have with LLMs?
It’s speed. Grunt work & time gets in the way of seeing out ideas and making them realities. You can test ideas significantly faster and get to answers much quicker so you’re moving mountains instead of pebbles.
If I’m working on something that is properly documented, it’s a waste of time to ask a human or search for something when a machine can find it in seconds so I can continue my work instead of spinning in circles for a hour.
Why am I as a single dad having to spend an entire night working on a meal plan for the week when I can have AI take my parameters, point it to my sources and drop me the plan every Sunday at 5pm without paying a service to do it?
Why should I look through articles and articles of local events when I can have my local AI ingest everything and bubble up what’s important?
I can’t trust news that pops up on Reddit by bots or government controlled media outlets. But I also don’t sit around on my ass all day to spend hours on news. Why not have my local models point at places like AP and Reuters and pull down all the articles in the day and summarize it for me for a 15 min read?
The internet has been enshitified, if I want to learn about something and research whatever, I have to dig through layers of sponsored content, click bait, and straight up lies. Why not build a RAG pointed at places I can trust and ingest data dumps from highly respected scientific and medical studies so I can get -real- answers?
I’m a high output army of 1 with responsibilities taller than me. My greatest enemy is time and I’m regularly trying to shove 35 hours into a 24 hour day.
AI has been a gift.
But you need to do it local and design your own stuff.
