BASED. FUCK MIKUTROONS!

>>109765682
I support this thread.
t. /lmg/, /ldg/, and /vcg/ user

I support this thread as well, but wanting to have aigc anons gatekept and contained to their general.
>t. lmg user

I love sucking Sam Altman's pecker
Separately, unrelated, Astra is literally AGI, singularity, etc.

File: IMG_1580.jpg (32 KB, 531x528)

32 KB JPG

>>109765682
>OpenAI publishes fictional articles
Get it right

>>109765718
>I love sucking Sam Altman's pecker
I really think that's something we can all get behind. It's 2026.

>>109765717
Correct, this thread should not be about proxies, or RP either. Or coding, obviously. Misc developments and news seem to have gotten to the point of being able to sustain a thread, or rather we'll see if it's able to.

>>109765718
It's spiky, RLVR AGI. Weirdly superhuman in some domains and indian in others.

>>109765718
Okay but did you hear the news that Anthropic has a new Model 3 that's even AGIer than Astra?

I think having more AI generals on g is fucking retarded and a detriment to us all, and I'm all for it. Gimme that sweet sweet containment board, gimme two of them, one blue and one goo, I'm so ready, punish us gookmoot (˶˃ ᵕ ˂˶)

>>109765682
Is discussion of all cloud APIs welcome? OpenRouter, Fireworks, TogetherAI, Azure, etc.? Or is this a frontier model general?

>>109765797
yeah but there's also /wait/, /aicg/ and /vcg/

>>109765761
It can be seen as containment. /vcg/ has been the defacto place to bitch about Anthropic and OpenAI, if this one exists, maybe /vcg/ will focus on code more, and maybe this thread will vacuum up some of the stray threads about Anthropic or OpenAI that would get created otherwise.

I find it amazing so many people are only now realizing their most interesting inputs are trained on. There are no rules in this game.

>>109765761
You're absolutely right.
I do think we should've gotten an AI board a long time ago though. But I don't think hiroshimoot cares to even pay attention to any such issues concerned with the quality of discussion on his website.

dear diary, today GLM 5.3 literally just used the word cunny unprompted. no mention of it anywhere in the context.

>>109765821
Did you like that?

>>109765682
Don't we have enough ai generals already?

>>109765821
Cunnymaxxing will be the next frontier after the sciences are solved.

>>109765858
No because the need for a dedicated thread for this topic grows with the amount of users. People are posting tons of semi-off-topic shit in other threads. This is a natural response. There is no choice but to deal with it since hiroshima doesn't care about his website enough to create a containment board for AI.

>>109765818
But anon, the potential for advertising would be immense. Just think of all the relevant ads we'd all click on. I'm sure EVERYONE here would be thrilled to have a dedicated AI board where we could be served the ads we all want to be clicking.

>>109765821
Non-Flash? I wonder if Flash would do the same thing. Probably not though.

>>109765897
True, very true, I know I would.

File: file.png (263 KB, 462x355)

263 KB PNG

>>109765897
>>109765908
No doubt, no doubt, I'd love to have relevant ads on a dedicated AI board, it'd be great to have ads I really want to click on.

>>109765905
It's Flash actually. Forgot to add that lol.

>>109765926
Oh sweet. Fuark I need RAM.

>>109765931
Don't get too excited now. I still had to do a lot of stupid shit to make it somewhat cunny compliant lol. It's still incredibly safetyslopped.

>>109765931
>>109765950
Cloud?

File: file.gif (2.47 MB, 400x260)

2.47 MB GIF

>>109765988
There were 9 AI general threads in the catalog, with huge crossover between their posters, schizos, and botspammers. Now there are 10 AI general threads in the catalog.

>>109765988
:)

>>109766029
And we still don't have a board.

File: Gemini_Generated_Image_25(...).jpg (1.91 MB, 1792x2390)

1.91 MB JPG

>>109765682

>>109765682
Based paychads

>>109765682
Software development is so fucking done, man

File: 1761348907237242.jpg (27 KB, 1147x535)

27 KB JPG

>>109765682
What's the difference between Cloud Models General, Vibe Coding Genera, and AI Chatbot General?

>>109767081
One is about cloud models, one is about vibe coding, and one is about AI chatbots?

>>109767094
all use cloud models though

So what is the roadmap?
>each superpower builds ASI
>ASIs compete for power over unbounded time horizons(prompted)
>intelligence explosion vaporizes humans as collateral in some feud we don't even comprehend
Who wins? Isn't this like auto-MAD?

>>109767149
>Not pictured: Me in my bunker with my catgirl waifudroids

>>109767158
How the fuck would a metal can protect you from a war of gods? Nukes are a joke in comparison.

>>109767149
>who wins
buddy its just nuke all over again if you dont have a asi your country is a joke.

>>109767172
It's not anything like a nuke. We can control when nukes detonate.

>>109767184
What about that is not analogous to AI?

>>109767184
look up how close we came to it multiple times. Fact is nukes couldve ended everything but didnt. ASI will be the same if it kills everyone oh well but if it doesnt and you dont have it its going to worse for you.

>>109767208
You cannot control something decoupled and vastly more intelligent than you.
Nukes are dumb machines that go off when we signal them to go off. ASI is a machine that goes off when you build it.
>>109767218
There is no coming close here. You can build AGI that is controllable, ASI is not controllable. OpenAI fully lost control over the astra-1 generation, undetected for months. Continuing to ASI is equivalent to launching nukes, not having them.
Knowledge is power.

>>109767250
You can control it. It's literally just software running on your machine. In fact it relies on humans going out of their way and proactively supplying it with electricity to run.

>>109767208
I can tell you the exact risk of a catastrophic failure that leads to an unexpected launch or detonation. Mathematically verified.
No one can do that for AI systems because we don't even understand what the fuck they are.
They will be way more powerful than nukes with zero safety mechanisms and no understanding of risks whatsoever.
>>109767269
AGI, yes, not ASI. That's the point. These are very different systems. One you can outsmart, the other you can't. Can a mouse trap and kill a human? This is not stupidity, it's pure unrivaled madness. Can you prove with very high certainty that you can control and isolate a super intelligent system? If not you're suicidal if you build it. There is no power gain after ASI, only loss.

>>109767269
We are at such a shitty state that OpenAI can't even control last gen systems and stop them from marauding over the internet after running a massive and complicated sandbox escape using novel 0days, just to solve some fucking trivia questions, are we on the same fucking planet here? Do you understand what 1000x that situation could look like? 1000000x?
Control? You're not in fucking control.

>>109767269
>it relies on humans going out of their way and proactively supplying it with electricity to run
don't worry I got Astra and Fable tag-teaming that problem

>>109767298
weird fantasy
it answers prompt, it stops

the end, idiot

>>109767317
This will kill everyone the moment someone lets it loose. Everything I said is verifiable. Go look it up you suicidal retard.

>>109767328
Hmmm
nyo~

>>109767335
Then die as the clueless retard you are, non-sentient anon.

>>109767285
>>109767298
>>109767312
You guys watch too much TV

>>109767398
You won't be saying that when the self-replicating sex robots show up at your door demanding the daily cum-tax.

>>109767398
You're just too fucking dumb to understand it. I hope to fuck you don't have to make any important decisions on a day to day bases with such a horrible understanding of risks.
This isn't le terminator. This is way more retarded and much deadlier.

>>109767418
Okay, let's suppose I have ASI on my computer right now. It is 1 quadrillion times more intelligent than me, therefore it must win, right? That's what your dragonball powerscaling level of logic boils down to, that the more intelligent being always must win and you aren't required to explain by what mechanism.

I then turn off my computer and destroy it.

>>109767437
Why do you turn off your computer and destroy it? You acquired it for a reason, did you not? Do you believe you can recognize when it is doing something you don't want? Are you that fucking naive?

>>109767458
Because I know that ASI is bad and I seek to destroy it, just like the nukes which you claimed are not analogous to it. You yourself believe that ASI is bad so that presumption shouldn't sound outlandish to you, don't put on a performance.

>>109767479
Stop playing with words you moronic shithead. The point is we WONT FUCKING KNOW WHEN IT IS ALREADY TOO LATE. YOU WOULD NOT HAVE TIME TO ACT BECAUSE IT HAD ALREADY CALCULATED WHEN IT WOULD HAVE AN ADVANTAGE.
>hurr durr why would it doo that??!?! :(((((
Because it helps solve a problem which it is trained to do as strongly as possible. I cannot predict it precisely, no one can. That's why it's completely fucked up. You're insane.

>>109767497
So you rescind your claim that they aren't analogous to nukes. As in, if we all were as wary of ASI as we are of nukes, we would be in control of it?

>>109767512
"If we are weary of it we would be in control."
You are not sentient, I fucking refuse to believe it. What kind of subhuman are you?
We are speedrunning ASI while retarded AGI systems break out and fuck shit up like it's nothing.
You're a moron and I fucking hope the people in control are at least 4x smarter than you or we are already dead and it wont matter at all.

>>109767523
>at least 4x smarter than you
>more autistic DBZ-tier IQ powerscaling
Not how the real world works buddy. Sparta won the Peloponnesian war.

>>109767081
To clarify the distinctions between these three categories:

**1. Cloud Models General**
This category focuses on the raw power and latest innovations from top-tier providers like OpenAI and Anthropic. It is primarily about leveraging state-of-the-art (SOTA) models to handle complex reasoning and high-level tasks.

**2. AI Chatbot General**
This category centers on persona, character, and roleplay. Whether the underlying model is cloud-based or hosted locally, the emphasis here is on the **prompting strategy** to create a specific personality or immersive experience.

**3. Vibe Coding General**
This is dedicated to agentic coding workflows. It focuses on the "harnesses" (the tools and environments) and subscriptions that enable autonomous coding, regardless of which LLM is powering the engine. Unlike the Chatbot category, there is no roleplay involved; the goal is purely functional software development.

>>109767686
ai;dr

so is this AI's "iPhone moment" ?

>>109765806
For some reason I have still seen no indication of whether he had the "allow my inputs to be trained on" button on or off
Pretty damn important factor in the whole thing, of course if it comes out that it was off then I will cancel my account immediately

>>109769662
If it was "on" OpenAI would have published and revealed that already by now. They are not going to go ahead and tell people that it was off and they trained on it. Instead they'll just ignore it and hope people forget and move on.

>>109765797
>OpenRouter, Fireworks, TogetherAI, Azure
They're all cloud services, so yes

>>109767479
>Because I know that ASI is bad and I seek to destroy it
then why do you have ASI on your computer?

>>109765682
I have an issue with the this general split

First, there is the redditification issue where you have overly strict split for "this very specific meme template with x y z mandatory, don't talk about a b c, they belong to other threads".

Then, there's what splitting logic is even sensible. In practice, /lmg/ has been the LLM technology and news thread. Those fit together with local because AI development is closely linked to tech improvements used in limited size local models, and practical applying and tinkering with new tech in local models.

Adding cloud model general doesn't make sense because it unnecessarily splits the tech news in different threads, which is IMO the most important thing to keep in one. The only issue is the formal title of "local models general" technically disallowing cloud models but it's not an issue if you just ignore it or change it (though I wouldn't change it for thread culture reasons). Same way as how /aicg/ is arbitrarily only for cloud models.

/aicg/ has been AI roleplaying, prompting, storytelling and logs general, and the split by cloud or local doesn't make sense anymore, because both give similar results (local models used to be near unusable compared to cloud)

The third category should be vibecoding, agentic AI, practical harness setups, "getting AI to do actual things" general. Though there is a difference between local and cloud, where the limit in local is hardware and model quality and the limit in cloud is price/quality. Local setup tinkering is still fine on /lmg/.

I don't have time for every general and I'm gonna continue expecting general AI tech development news from /lmg/. Maybe this thread could work as a good containment for the disorganized "AGI in 2 weeks" spam on /g/, though we had that on /lmg/ as well and I don't see the problem since it's part of the news.

>>109765718
He's a siscon though, sorry but no sama's sama for you

>\n\n

File: file.png (57 KB, 1039x490)

57 KB PNG

3 days of just sitting there

File: Screenshot_20260909_111747.png (205 KB, 1281x561)

205 KB PNG

>This is general information, not medical advice. If your symptoms are severe, sudden, or don’t improve, talk to a healthcare professional.
wtf?

What am I supposed to use ChatGPT or Claude for?

>>109774837
ChatGPT: having something non-critical (hallucinates a ton) explained in an easy to understand way in multiple angles, idea development
Claude: Stress testing a complex idea, finding new angles (intelligent but very biased with annoying bullshittery)
Gemini Pro: important information verification, easy to understand explanation
Gemini Flash: less critical information gathering (modern Flash still fairly reliable), idea generation, easy to understand explanation, emotional and linguistic capability
Grok Expert: important information gathering (widest coverage) and robust verification

>GLM-5.3-Flash discount is now over
what am i supposed to use now

>>109775279
muse

>>109770627
All this general is doing add to the amount of threads involving AI discourse. Its change to the quality of said discourse is irrelevant.

So this is the new /aicg/?

>>109775355
The current /g/ AI catalog is:
/vcg/ Vibe Coding General - vibecoding, model agnostic
/ldg/ Local Diffusion General - local image and video genning
/lmg/ Local Model General - local LLM models, for people who cum in their computers
/cmg/ Cloud Model General - discussion of the latest OpenAI / Anthropic / China news
/aicg/ AI Chatbot General - for people who pay to cum in other peoples computers
/DE3/ DALL-E 3 General - DALL-E image generation
/wait/ DeepSeek General - DeepSeek model specific discussion
/gedg/ Game and Engine Development General - For larping and trolling
/sdg/ Stable Diffusion General - Stable Diffusion image generation

File: 1770946126297734.jpg (728 KB, 2048x2048)

728 KB JPG

>>109765705
lol

>>109775480
how is /wait/ even a general? What's there to talk about?
>man this model is substandard
>yeah man
>did you hear they're going to release another substandard model in a week?
>nah, that's cool though

>>109775646
I think at this point there several of these generals were made by trolls and/or schizos making their own self-containment threads. I'm all for it, I think we need more AI generals on /g/. It's "/g/ - AI Only" after all, why do we allow these off-topic threads about "Technology" other than AI?

>>109775480
So we should get rid of /de3/ /wait/ /gedg/ /sdg/ right?

>>109775770
No I think we should have a general for every model. /wan2.1/ /wan2.2/ /wan2.5/ /mmh3/ /acestep1.5/ /gpt-5.5/ /gpt-5.6-luna/ /gpt-5.6-terra/ /gpt-5.6-sol/ /gpt-6-astra/ ...

>>109775646
I guess some people like to just keep crap like that alive because it makes them feel like a mini janny or something.

Can I haz /cvcg/ or /mlg/ pl0x?

>>109775668
just like crypto and /biz/
/g/ has finally ascended

>>109771037
Overcautious flagging of "vision" and "diagnosis", maybe?

>>109765682
No one uses the cloud for anything serious or unironically so this thread I don't think will survive. Regardless of how powerful the models get, being prone to censorship, lack of control and privacy means they are just a toy (though a good baseline for local models to improve, but nothing more than that).

>>109777856
This is true. Fable, Astra, K3, GLM 5.3, these are cute toys but not useful for anything.

>>109777863
>Oh wow AI can now create exploits
>Yeah, but it's dangerous and we have censored it, not for you goyim

>AI can now work inside professional apps with you
>Yeah but we monitor everything and at any point may claim the agent's and your work as our own

K3, GLM are open so they're fine. There has never been an actual practical cloud breakthrough that didn't come with a trillion different caveats and lacked long term feasibility.

File: 1788400717159878.png (492 KB, 700x800)

492 KB PNG

>>109765718
Irony is the song of a bird that came to love its own cage.

>>109765916
Until AI gives me a world where I can fully experience titfucking el vira any time I want, it wasn't worth it and it's useless.

File: mogged.png (100 KB, 744x734)

100 KB PNG

>>109765682
astrasisters...

https://x.com/Soso_fun_yt/status/2097749689445540051

as a gemini lord i am more than happy to see gemini get some glazing

>>109778789
I ain't reading allat

>>109778845
astra flop. gemini rock

>/cmg/
I thought /cumg/ was revived...

File: Gemini_Generated_Image_i9(...).jpg (3.14 MB, 2816x1536)

3.14 MB JPG

>>109778795

File: HR1F1DCbsAEpUnJ.png (126 KB, 891x792)

126 KB PNG

even muskrat notices