you faggots lied to me, especially that one schizo retard who forced snailcat meme. this shit sucks ass so fucking much. it maybe can impress twitter trannies, but the moment you actually start using it for real projects it shits itself immediately. it uses 68% of 5 hour limit tokens just to put shit in its pants.
>>109106514pretty much lol, proper ai uses and productivity gains are much more limited compared to what's being shilled. It's all astroturfing
>>109106525When astroturfing ends it will become too expensive to even use properly.
>>109106596no doubt. It will be funny watching retards cry how they can't slop without paying a fortune lol
>>109106643What even are local models.The slop is here to stay.
>>109106691i don't think the tards who shill here know how to set them up.
>>109106514Skill issue. It's neither a magical panacea, nor utter garbage. It has its uses.But seriously, technologically speaking, why would bids prefer wires over perfectly good trees??
>>109106855I was promised a revolutionary technology, instead it uses MY WHOLE DAY LIMIT OF TOKENS to fix html/css
>>109106514okay but why are you asking about bird welfare, is it some race joke
>>109106514The fuck did you expect? The whole thing is run by a bunch of grifters.>>109106886It cannot into html+css because those are too complex. Produces write-only slop. It may seem simple to you, but LLMs cannot write html+css reliably at all.
>>109106886look, if you spend around half a million dollerinos and run 300 agents on loop, it might finish html/css (eventually?)
>>109106941then how does it supposedly handle much more complex stuff effortlessly? at least according to corporate shills and enthusiasts. is it all just gaslighting? i feel like at some point thing plateaued and everyone pretends it didn't
>>109107249a lot of stuff it's marketed doing is very simple but to layman it looks like magic. Like when everyone shilled how fable or whatever one shot a minecraft, when in reality it just cloned and slopped up existing project on github.For other complex stuff that ai sometimes manages to do it depends if exact use case is somewhere in its training data. You're basically googling in a compressed database among a bunch of github repos and other shit.Even then it will product prototype-tier crap.Complex shit that actually works isn't effortless at all and requires a lot of work and experience. Vibecoding is a trap for morons.
I dunno, I had to ask an LLM today on how to clean an old motorcycle carb and it was great, saved me hours of researching. Last time I had to do it was more than 10 years ago, and I have no recollection of what I did.
>>109106514If you had done a bit of research you would have known Anthropics $20 plan is unusable
>>109106514If you buy a circular saw and don’t know what you’re doing you’re going to waste your money, ruin your project, and maybe even cut off your thumb. If you pick up a new tool you’re not going to be effective with it. Blaming the tool is just retarded.
>>109106941This guy actually thinks HTML+ CSS is complex my brother under Allah we were writing that shit as children for MySpace
>>109108039>buy a saw>try to cut something very simple>battery dies in 5 seconds, saw's handle breaks in half>wtf this product is shit>erm, actually, don't blame the tool
>>109106514Literally skill issue, trans. We vibeGODS won
>>109108129Let us know when ChatGPT breaks in half for you
>>109106514opus is for Americans. We are just better than the rest of the world. God gave us Fable, because he trusts us to do good for all of mankind. 3000 years ago God promised the white people Fable.
>>109106525> mfw i realize the "ai revolution" is just glorified autocomplete and a hype bubble
>>109108370herbs.it's not.
>>109106514vibeGODS will be victorious against filthy hordes.
>>109106643> paying a fortune for slopyou mean the 20 bucks a month that's less than a single takeout meal? kek
>>109108426That Anon implied that post-astroturf quality/price will drop
>>109108134Still waiting.
>>109106707To get a decent local model running you need more than a single powerful PC. At least like 6 mac minis put together or a GPU cluster. And it probably still won't be comparable to something like Sonnet.
>>109106514Very useful for quick prototyping and automated testing/debugging.Retarded for actually building a somewhat complex system
>>109106514Well, /g/?
>>109106707Even then, there always will be service hosting them for retards.
>>109109409I think you are really outdated.You can run something better than sonnet for less than 10k in hardware.And either way >>109109925
>>109109366i feel like this is the best counter for those 'vibegod' retardsliterally nobody who are using ai in any meaningful way wants to dox themselves nor even acknowledge that they are using ai publicly over such a retarded and insufferable meme let alone really agree with what such a meme implies
>>109108246The analogy was retarded, was his point, and he is correct. AI isn't properly considered a mere tool.
>>109108417aren't the cats fighing the machines? lmao, vibefags have their brains so cooked they can't parse the images they gen
>>109108426even the 20/mo is very limited. in my company we pay per token and the bill for june (massive price hike) is expected to be around 5-10k per developper
>>109109927show me the under 10k hardware that can run a 1.5T model
>>109107322this, it's basically a better search engine, but every time you get off the guardrails and ask about something not in training data, it hallucinates.thankfully, most developers are just making a copy of what their competitors are doing, so models are likely to be useful since they've already seen the code.I asked fable about a problem I had with a very recent library, it was able to search internet for the documentation, but only outputted nonsense
>>109106514>>109109477I love this floral guy.
>>109110181what is the origin of this critter?i see it on twitter sometimes and everytime i wonder
>>109110082ikr. Should be fighting swarms of filthy humans.
>>109110082i dont think those 'vibegod' slop images are made by someone who really supports ai development but someone trying to play the 4d chess by being obnoxious on purpose
>>109106707it's not that, the problem is lack of hardware availability and affordability
>>109106643>they can't slop without paying a fortuneI'm already switching over to all-Chinese models, have fun with your gated burgerslop
>>109108043I guess you have never written any modern CSS except your myspace background-color: red; shit
MCP + AGENTIC AI + SKILLS -> Learn these.1. Google BigQuery - the absolute single most important thing here.2. https://cdn.openai.com/business-guides-and-resources/a-practical-guide-to-building-agents.pdf3. You need to learn how context is stored.4. https://repomix.com/ - You can now pack entire git repos into an XML document and one shot it at the LLM5. Pick ONE cloud SAAS and MASTER it. Really, the only two worth doing are Google Cloud Platform and Amazon Web Services; you will need EC2 anyways6. CONTEXT7 MCP SERVER IS MANDATORY IN ALL HARNESSES This is all you need to know about AI today
>>109111763>MCPpoint and laugh!
>>109110156glm 5.2 is 800B retard.And minimax m3 is 400BAt q4 you are looking at about 400GB for glm and 200GB for minimax.You can just get 3 to 6 r9700 (1200 each) and do the rest on ram inference.You can expect 20 to 60t/s with such a setup.And anyway, you can always run qwen 3.6 27b pretty cheaply and it's already a good model.And again, there will always be providers for open models if you can't even afford a poorfag setup.
>>109111763>bark bark barkGod, I love buzzword-spouting automatons.
>>109111763half of this shit was outdated 6 month ago lol
>>109110156>>109111960also if you want to be realy cheap but still fast you can get sxm2 chips, they are at like 700$ for 32GB vram and like 200$ for 16GB vram, about 900GB/s.
