how do you respond without sounding like an AI cultist?
why respond at all
There are many things that can enrich one's life, wasting time on pointless arguments on Twitter is probably not one of them.
>>110279018>>110279022concession accepted
>>110278988It's called SI
>>110278988He's right, though. Not only that, whatever "skill" needed for AI becomes irrelevant 4 months later, there is no point in keeping up with the mania, just do whatever you want.
>>110278988Learn to become better at writing to express your intent clearly and thoroughly without rambling. It\s a useful skill even outside AI, for example, to communicate with coworkers efficiently. Modern models are good at inferring what you meant even if you write some shitty 4 word prompt, but the more context your provide, the better the result will solve your problem without doing shit you never needed.
>>110280344weird, because AI shills seem to be exceptionally bad at exactly that.I still have to find out what the>adapt AI or get left behindeven means.And the dude in OP also didn't write>learn how to express yourselfIt seems like there is some kind of:>my AI only works if we all use itlogic going on here, which seems similar to the>my vax only works if you get vaxxed as wellfrom a few years ago.Otherwise, whats with that aggressive nondescriptive shilling?
>>110280402Yeah, I agree with that, actually. That's why they believe there's a real skill to "prompting" in the first place. Because they are absolute fucking dogshit at communicating their intent and noticed their results got better following prompt templates someone else made. All the model needs from you is context on what it needs to do. If you're able to provide that, congratulations, you are an elite professional proompter.I fucking hate this technology, but I have to live with it.
>>110278988There’s still something to learn because it’s still nowhere close to AGI level.
>>110280344Nope you don't even have to do that. LLMs are very good at understanding what you mean even if you are incredibly inarticulate.
>>110280344There's more AI specific thing. Learning to minimize token costs by things like maximizing cache hits, knowing when to restart context, knowing when to subagentify, etc. AI needs to be learned to be used effectively.
>>110280441It sounds like you could easily learn every single one of those in a 2 hour course. It doesn't even remotely compare to the knowledge you needed for a real job where you would acquire knowledge in practice over years and still have new things to learn for as long as you live.
>>110279032SI SEÑOR
I just say to the SI "rewrite this prompt like I'm a 200 IQ super genius. don't make any mistakes".
>>110278988He's right. No need to "learn" how to type a message into a text box and click send. Especially when using AI means you start relying on it and your brain atrophies. Just keep coding normally. The way I see it, if it is a bubble, you'll be months or years ahead of all the people who have to learn to code for real again when it pops and if it isn't a bubble, you're only 5 minutes behind people who've used AI all this time.
>>110280434As much as I don't really like AI, I've recently started utilising it to help proof-read my shell scripts and look for ways to improve upon them. I don't see the point in refusing to use a technology just because the inventors of said technology have questionable morals.I gave it a shell script that I wrote a while back (no AI generated code. I wrote the script myself, like a "snailcat"). I was surprised by how it was able to immediately infer why I put certain commands in some of the functions, despite not leaving any comments explaining their purpose, nor providing any direct indication in any of my prompts as to why I put those commands in there.It just...read my code and correctly deduced the reasons for those commands being included in my script, despite the inclusion of some of those commands being a bit ambiguous. A human would probably make the same deductions, but I was surprised that AI could make those deductions.Interesting stuff. I don't think I'll get it to generate code for me, because I enjoy writing code. But it might be a useful tool for proof-reading code and suggesting improvements.
>>110280323>Not only that, whatever "skill" needed for AI becomes irrelevant 4 months later, there is no point in keeping up with the mania, just do whatever you want.>>110280453>It doesn't even remotely compare to the knowledge you needed for a real job where you would acquire knowledge in practice over years and still have new things to learn for as long as you live.
>>110278988Communicating with an LLM isn't the same as communicating with a person. It has no personality, background, or anything of the sort. And it sucks at inferring things from what you say. So you have to actually know something of what you're talking about to get it to do what you want.
results do not depend on your skills, but rather on the quality of the model. Thats why there isnt any skill involvement. Unless by skill u mean watching few tutorials on yt about how to setup best models for particular aplication
>>110278988His analogy is correct. AI is an automation tool used to automate away thinking and programming. There is no skill involved in using it.
>>110280535how are they wrong? you seem to be coping
>>110280690>"[...] whatever skill needed for ai becomes irrelevant 4 months later [so using ai is not an intellectually valid job or hobby]">"[...] you would acquire knowledge in practice over years and still have new things to learn, [so using ai is not an intellectually valid job or hobby]"which one is right cretin
>>110278988
>>110280705>inability to understand arguments>frog postingchecks out
>>110280727>can only type in memes he's learned from my posts here over years instead of explaining rhetoric you must be one of those "ai users"... computer, do!
>>110280705Both. They aren't mutually exclusive. Just because how agents are managed is evolving doesn't change the fact that they are still low skill while most crafts are deep enough to last for a lifetime of self improvement.
>>110280780>factbelief.
>>110278988what has the guy at the top made with ai that was one shot without any prior experience?
>>110278988I can't.
/g/ always underestimates how severely retarded normal people are, it will never ever be "Anyone can make software" because people are too fucking stupid to do so even with a magical coding toolI work for a Fortune 500 company and you legit would not believe how dumb people who work here are, and they mostly earn over 100k$ a year.
>>110278988It's true there's nothing to learn. It's also true programmers, artists, mathematicians (and soon musicians, physicist, financefags and a whole swath of white collar workers) are obsolete and their skillet is not needed anymore.
>>110280715:)
>>110280476This. My junior colleague keeps crying that hes afraid of being replaced and he literally, non metaphorically, can't do anything at all without ai, and he routinely fucks things up. He asks me what should he do to not get replaced and I told him what you said, or to just change career, but he keeps saying "yeah but why learn? I'm slower if I do it manually. AI is the future brooo" (he's slower because he can't write even a fucking for loop in python). He unironically thinks he can """learn""" ai, but despite him wasting time with the latest useless gimmick all the time, I'm still better than him at agentic """"engineering"""" because it's literally text prompts and markdown files all the way down
>>110281393wait, so you're saying that you can actually... learn... to get better at using AI? holy shamoley
>>110281449Nice strawman, 7/10
>>110279032Your called gay faggit of yore
I still think you can't really make good applications without knowing how to at least read code because AI makes a lot of dumb mistakes. But its pretty obvious thats only going to be true for another few months
>>110281577Who makes dumb mistakes? Opus? Only if you don't give it enough to go on man. If you actually take the time to make it research the issue, it fixes its own mistakes already. And yeah, this is likely the worst it'll ever get.
>>110278988“Learning AI” is advice to old experts not young newcomers. If you are some old expert who is two or three decades in their profession then you learning AI helps you use the AI with your accumulated experiences and contacts to babysit the AI (because despite how capable AI is, it’s still hallucinating in tasks that can’t be automatically checked with some software). It’s telling boomers and GenX to learn how to prompt, how agents work, that LLMs are really really smart next token predictors, what is context window and so on, something obvious to all the people on here that takes like a weekend for young mind to learn. If you are young and have no experiemces or history in certain job market then you are fucked, you have to compete with clankers that can and probably did all of your homework and thesis, everything readily available as learning material for you has been scraped for chatbot training data and Opus 5.5 probably already mogs you in these tasks.
>>110280434You absolutely do, the less ambiguity in your words the less likely it is to spend doing shit you didn't ask it to do, probably the easiest example of that is trying to generate a highly specific image
>>110280441Those are significantly easier than lots of things AI already can do, so these "skills" will also be useless within a few months.Or do you still put "trending on artstation" into image prompts?
>>110278988>engaging in any discussion on social cancer... why? retarded much?
>>110278988Using AI != using AI to create something good.To create something new you must have knowledge on the subject, AI has all human knowledge inside of it, but it can't create things by itself because it has no reason to do so. That's why the user having the knowledge itself is crucial to create something good. Once the model spits out content the user can add more context of what it wants and the AI will improve the content, but it will never create stuff by itself. Sure it can infer many things nowadays and this make them better than older models by default, but a human still needs to give the AI directions on what the user actually wants (the FINAL product)."Get with the times, use AI" is much better to say, but you still have to learn things.(fuck im really getting dumber, took me like 20 minutes to write this shit post)
>>110278988This is AI.You make a wish, and your wish comes all fucked up if you don't think on every case and corner and describe it EXACTLY how you want.99% of people using AI are making retarded wishes.
AI is here to stay. Simple as. It is a force multiplier. It allows people to accomplish more than they were before. Its simpler than google searches. Anyone who is against AI is just a luddite. Its a better tool. That's all.The Jewish banks are insolvent. The system is going to crash. They will blame it on AI, but its fractional reserve lending is the cause. We told you fucks this was gonna happen. That you would be made obsolete. And yes. it takes some effort and technical knowledge to use AI. My stupid parents who can't even use their phones correctly will never be able to leverage AI. But you dumbass millenials who suck jewish dick will be able to.
>>110278988He's right that "learning AI" is useful, but in the sense that leaning how to fine tune and make your own harness is extremely valuable. However that's not what he's talking about.Most people shilling AI who are saying you're going to be left behind are not doing either of those things. They're retards who are still basically believing that they're prompt engineering in 2026. Funnily enough, they'll also argue to the ends of the earth against learning the more technical side of AI.It's very transparent that this is just retards who are norm policing. These people have no idea what the fuck they're doing. They're just upset when you do something that even slightly hints at the idea that they're being taken for a ride.No you will not be left behind. At the level in which these mongoloids are using it, the guy replying is right. There is nothing to learn.
>>110278988most non-technical people don't even realize you can use harnesses to direct your computerlike you can 10x your ability to automate shit with that alone.
>>110282559What do you know about the Luddites?
>>110282873They're people that hate jeets and don't know yet.
>>110278988There are things to learn but it's on the same level as learning a coworker's personality and competences. Different models have different quirks, strengths and weaknesses.
>>110278988H
>>110281356Yet they are still employed
>>110283577Not for long, they are slowly but surely getting replaced.
>>110278988>(are people selling pizza ordering courses)found the marketer or the PDF FOMO consoomer
>learn AI use>next model can one shot your week long project with far less instruction and wrangling >repeat forever
>>110282770>leaning how to fine tune and make your own harness is extremely valuableNo it's not. Just ask Claude to do it.
>>110278988You're right, Claude made my level editor for $5 this week and all I did was boomerprompt with a few more technical requests, no code, I haven't seen a line of code for this whole project so far
>>110283586Who's coping now? What's your evidence?
>>110279032THANK YOU
>>110278988>2-digit IQ normgroid needs to learn how to writengmi
>>110283653Yeah until the VC money dries up and that week long project suddenly costs $25,000 to one shot
>>110283653"Draw the cover art for a game named mega man, it is a blue robot that shoots and jump on platforms, you have one night".
>>110283816Dammit claude you did it again!!
>>110283662Kind of.You CAN just point the LLM to an existing harness and say "copy that", but one of the more promising aspects of AI is that you can use programs like the Pi terminal to extend themselves, e.g. I had it write an input server so it could test a project of mine "headless" as well as the .md spec and JS it would need to do the job consistently. I wouldn't really have thought about this before, but making a way to pipe sequences of commands + waits to a server is a pretty interesting pattern and is kind of reminiscent of Change Data Capture.So, some people right now are using AI to effectively create all the niche tools they wouldn't have tried to make otherwise, like artists creating unique brushes for digital painting or video editors finding better ways to do NLE.Essentially the one area AI isn't causing cognitive atrophy is people using it to make tools.>>110283653Do we have any "Trust me bro" benchmarks about code quality/code volume? I'd be interested if someone was trying measure how maintainable Agentic projects actually are.
>>110283832It's a much more advanced model named human being.Yet it still failed at the task because prompting itself is defective."One shotting" is basically "just fill everything with slop", and it will be the most generic shit possible.If you want to make a good game or software, you have to spend a few months, AI or no AI, even if you spend these months prototyping shit and drawing detailed plans and going back and forth.It was never ready on your head.
>>110283814two more weeks
>>110284410Anthropic is junk stock, it's over
>>110283662You may as well just ask Claude to make you money.
>>110283839No. Most research on agent AI projects measures time to completion as the success metric. There's actually a lot of blindspots in the literature. How would they measure the robustness of outcome anyways? We need more time to start looking at survivorship, and even then I expect those studies to be a little bit flawed. Like I imagine most AI startups are going to burn out like regular startups do, and there's a confounding variable there that I doubt will get accounted for (startups burn out at extremely high rates)
>>110284747True, measuring the long-term viability of code bases was, and remains, a hard problem.It also doesn't help that there's a lot of programmers who have these stupid, secret rules about what makes code "good" that are just vibes and culture-based like Clean Code stuff.Code base size is an appealing metric for me, because the amount of code generally correlates to complexity and the ability of someone to understand the project, but then you still have corner cases like when developers play "code golf" or when sometimes not generalizing code is the right decision because it might scale to a nightmare OOP taxonomy.Your read about product survivorship is probably the right approach, since the market will decide in any case.
>>110280780But you said we don’t need skills anymore. How are you this dumb?
