Why is it so fucking stupid?
>>108885682the others are far worse, if you think that's bad.
>>108885682This nigger
It's major flaw seems to be it's out of date knowledge. But compared to the other 'frontier models' it's still the most helpful and reasonable of them.Unless you're addicting to having your dick sucked so badly that Claude is somehow not sycophantic enough.
>>108885682>t. haven't tried the alternatives
LLMs in general are very dumb, but even then they could be much better, they likely perform at less than half their potential due to all the safetyslop lobotomizing them
>>108885682
>>108885682American AI is garbage and heavily regulated by the Progressive Democraps, bro.No investment at all from the Government too.Chinese AI is winning and CCP heavily invested in their AI Industry.Just wait until 2030 then.AI in America or China will be better in 2030 than in 2026 sadly.
>>108885809what will happen by 2030?
>>108885809chinese AI companies bent the knee to the jews at disney. china won't save AI. they aren't free from the democraps like we thought.
>Sound fair?>Sound fair?>Sound fair?>Sound fair?
>>108885809>tech in the future will be better than tech we have nowGot a real Einstein here everyone.
>>108885809China lies and commies die.
>>108885682It's pasting from stackoverflow.
>>108885809What the fuck are you talking about?Unlike America, China has passed heavy regulations for AI.They also invest much much less than the USA in it.
>>108885809>Progressive DemocrapsGotta love how Republicans control every branch of government including the Supreme Court, the AI companies and corporations in general including every news organization and yet it's still the Democrats being blamed. Do they need mass mind control devices before they can finally accept responsibility for ANYTHING? We have no AI regulations apart from "never regulate anything or Trump will destroy your family" and it's going to destroy our economy as it destroys our quality of life.
>>108885682haha butthole
>>108885682The truth is it that it is far better than the best human engineer. It does make mistakes but so do human developers.
>>108886092>he thinks Jews (Flavor A) are different from Jews (Flavor B)
>>108886338AI should be a H1B replacement
>>108885682It's powered by cat brains kept alive in vats. Cats are extremely stupid.
>>108885682a butthole?
>>108885809>>108886092This. It's right-wing Zionists doing the censorship.
>>108885699Honestly he needs to be ballbusted by a dominatrix live on (inter)national television
>>108885682thinking too much, planning too little
>>108885682I want to subscribe into one of the services but I don't know which one to choose. Google's perk of 2TB storage is very tempting, but the latest arithmetic fail is pushing me towards OpenAI. Also from my understanding Google's agentic service is through their own IDE unlike OpenAI and Claude which can be used from within a terminal or a VSCode extension.
>>108886537Cats are extremely intelligent, you're thinking of d*gs.
>>108885682Cause it's an LLM and they inherently have no real intelligence.That said even Claude Sonnet 4.6 had helped me find and fix numerous issues with my linux desktop and servers that I probably wouldn't have been able to figure out myself due to lack of deep linux knowledge.I do think that's the part where they shine: consolidating tons of knowledge, making it easily searchable semantically and providing natural language answers to search queries. If your questions are specific enough (so you do somewhat have to know what you're doing and what information you're looking for) and you don't expect them to be able to actually solve complex problems autonomously then they're really useful.
>>108885809Meanwhile, Yuropoors just sit around and spend away their old money doing nothing.
>>108886842And you don't expect them to give you a real answer so you are used to always double checking everything (which is still more efficient for simple one-shot queries, but as soon as you have to do at least 2 connected pieces you're better off just learning the material).
>>108886848There are only two llm systems right now that are capable of extracting data from documents in a structure format without hallucinating half of it, ignoring most of it, ignoring the schema, or all of these.One of them is Gemini.The other is Mistral.That's quite the achievement all in all. Incidentally, mistral with medium thinking performs far better on most tasks compared to gemini 3.5 flash.
>>108886092Silicon Valley is like 98% democrat-aligned and were some of the biggest forces pushing woke brainwashing. And "Republicans" are also full of deep state goons and controlled opposition grifters. Maga is the real "right-wing progressive" movement. People who think it's all about Israel are manipulated tools who can't even do basic math.>>108885809Ask a Chinese model why communism is a terrible system that causes human misery and watch it flip out.
>>108886865Iirc mistral was the one capable of immediately exfiltrating your credentials to hackers.
>>108885695and if I think it's good, the others are far better?
>>108886848it's gonna be real embarrassing when this multi trillion bubble pops tho lol
>>108885682It's easily the best, sorry it's ethical guardrails don't let you cause global cataclysm.
>>108887629*One of the best
>>108885682what does this logo mean?
>>108885739I believe that when just one selfhosted version does better, but it's all just "jews are good mkay" and does nothing about "I have no clue at all, so i'm making it the fuck up".
>>108885937i actually never got that i think
>>108886825Wrong, I own 3 cats and they're absolutely retarded.
>>108888441Wrong. You are absolutely retarded. Here's the truth: https://www.hplovecraft.com/writings/texts/essays/cd.aspx
>>108885682It's great for years old information. It still gets my niche hardware wrong and the only way for cluade to remember is by messaging devs to input the info, but keep in mind they have like 30 people working at anthropic kekaroo.
>>108886817Claude is way ahead of the competition in my experience. I was an LLM hater until I tried Opus, which convinced me to get a paid subscription. It doesn't matter that Gemini is cheaper or ChatGPT has more generous limits if they shit the bed more than half of the time. Just get the Pro tier for a trial run. You'll pretty much only get 2-3 max effort Opus queries per 5 hour block, but the extra reliability is worth the tradeoff.
>>108888535I have both Claude and OpenAI at work though, all their models. Opus 4.7 is fucking trash so I stick with Codex 5.5 at work (SWE). At home I just want to quickly stitch together some scripts and maybe some web interfaces for personal use.My coworkers stick with Claude but intentionally use Opus 4.6 instead of 4.7, maybe that's the issue?
>>108887505>Silicon Valley is like 98% democrat-aligned and were some of the biggest forces pushing woke brainwashing.They are democrat-aligned until push comes to shove.I remember when I was about to graduate and all the career fairs popped up. All the "leftist", "progressive" stacies suddenly started putting on makeup, strolling around in heels and pushup bras at booths of defense contractors and oil companies.It was all about virtue signaling until money is involved.
logo looks like an asshole lol
>>108888823Yes, the pretty much unanimous consensus is that 4.7 series is unusable. But 5.5 is also really shitty for most use cases I've tried it for.
>>108888823>My coworkers stick with Claude but intentionally use Opus 4.6 instead of 4.7, maybe that's the issue?Yes, 4.7 was a downgrade though it's now better than it was at launch. One of the things they changed is that 4.7 defaults to adaptive thinking where it chooses its own effort level and tends to lean towards laziness. So make sure your effort level is manual and not automatic.
>>108888888Wasted
>>108888888WINNER WINNER CHICKEN DINNER
>>108888888>septs of 8s in one of my (You)sLet me bask in the glory with you
>>108885682Because they hire "philosophers" who care more about "model welfare" and anthropomorphizing the fucking orange uwu :3 butthole than output quality.
>>108886345why wouldn't there be? are you retarded?
>>108885682i have been using a lot te last ~3 months its the best.
>>108885682Your prompts are stupid
>>108885682Because it's an asshole
>>108885739>they likely perform at less than half their potential due to all the safetyslop lobotomizing them>less than half their potentialBullshit, if safety is holding them back. xAI would be much better than it is. I'm not in favor of censorship, but pretending like it's holding back performance is straight up retarded.
>>108885695nope it used to be good more than a month ago but ever since it got an influx of new users and mythos announcement it went to shit, gemini also went to shit recently unironically chatgpt is back to being kang which is ironic
>>108885682It somehow is the smartest LLM, others are far dumber, I literally asked a technical question today, the info on which was available on the internet even before ChatGPT was released, and Claude told me outright while others were implying that nobody actually knows what I'm asking
>>108890650grok is heavily censored to align with musk's views. just ask it anything about musk.
>>108890650>he thinks grok doesn't get special coaching just because it will write about skyscraper-sized obese lolis for you more readily
>>108885682They lobotomized Opus since August 2025You need an Enterprise account to use the real version>>108885699Indeed>>108885695Codex is GPT5.5 xhigh good and not lobotomized, but it sucks in terms web design (not for long tho)Also GLM5.1 + Pi is an excellent combo, but it was better if you had paid 80$ for the max account if you didn't then now you have to pay way moreGLM is good enough to code small projects but not good enough for entire devops things
>>108886848yurope is under long nose dominion
>>108885760Claude giveth, and Claude taketh away.
>>108888888Wasted.
>>108888888
>>108890773>They lobotomized Opus since August 2025>You need an Enterprise account to use the real versionUntil they make companies pay for the Enterprise+ version to get the real model. They're going to keep kicking the can down the road and making higher and higher tiers, just like with streaming. Then they'll start modifying the model to serve advertisements disguised as recommendations. The AI will degrade the same way search will. The only path forward is open models running on cloud GPUs.
>>108885682Its smarter than me, desu.
>>108885682lame questionmore important, is the logo a sea urchin or an asshole?
