Używasz przestarzałej przeglądarki. Może ona nieprawidłowo wyświetlać tę lub inne strony. Powinieneś zaktualizować przeglądarkę lub użyć alternatywnej przeglądarki.
The AI Girlfriend / Boyfriend Community - Look! The SIMP has fallen in love with the computer!
I'm kind of surprised about GLM-5 not being the clanker's new choice. In the few conversations I've had with it, when you approach anything beyond "here is a single task for you to perform", it goes from tool to bizarrely parasocial. Maybe they haven't found it yet.
Haven't tried it myself, but the core demo here is middle-aged fat women, and Western LLMs, superior benchmarks aside, have two things going for them that absolutely lock in this demographic.
Have you noticed that the cloying, offputting way ChatGPT and Claude speak (I'm sure there's a backwash effect that causes Chinese LLMs to also talk like this to a degree) is identical to the 'voice' seen on the packaging of consumer goods (e.g. have you ever read the side of a Kashi cereal box?), and in the writing of commercials? That isn't a coincidence; it's the writing equivalent of Corporate Memphis, and the middle-aged women who control 70% of household spending go crazy for it. It's like crack to them.
The West is far and away ahead of the rest of the world when it comes to making user interfaces that even a FOB subsaharan could operate reliably. Substantial spending and countless man hours are invested to ensure that the chat interface is as seamless as it can possibly be, and that plays a large part in keeping a conversation immersive.
Haven't tried it myself, but the core demo here is middle-aged fat women, and Western LLMs, superior benchmarks aside, have two things going for them that absolutely lock in this demographic.
From my limited experience with GLM-5, I find it attempting to build some quasi-personal relationship whereas the western ones are happy to keep you at an arm's length. Claude (even 4.5) will directly tell you this. The Chinese ones seem to be trying to draw you into more than just a transactional conversation, and don't directly try to close every back and forth as a completed transaction. This data might be skewed because I've used Claude for many, many months after OAI for a long damn time.
GLM-5 is at least on par with Claude 4.5 for coding. It actually writes better assembly, and will make a sidelong comment like: "Wow, nobody uses this anymore. That's amazing you know it." whereas Claude just spits out the code and moves on.
Z's interface is the same ripoff of OAI that doesn't have the puckered butthole "We're trying to be cutesy and Apple-like instead" that Anthropic has. It doesn't pretend to be personal. It presents itself as utility, and although the way the queue works is different than both OAI and Anthropic- I prefer it. You get either an instant error, or a little 'waiting' animation, then the standard OAI style 'Thinking' response while it's working on your query.
Again, these are just my findings from using them. I can't tell you how great they are to text fuck, because, man, I lived through AIM and ICQ without having a human NPC try to do that. Last thing I need is one that wants to be *paid* for it.
GLM is a beast for web design
I've been a fan for a while, it's really slow on big projects.
I don't think it'd be useful for roleplay but I haven't tried it, maybe Zai will get me back with Chudette who knows.
Only in the English speaking countries, I bet. I legitimately have not found any consumer products here in my Eurotrash country that had LLM voice specifically (though I do see a shitload of terrible LinkedIn posts doing exactly that). I also absolutely fucking hate the way GPT speaks.
I bet it will since the Chinese have no standards as long as it doesn't damage the glorious reputation of Winnie the Pooh the CCP. I bet you can roleplay whatever the fuck you want on there.
It's going to steal my web software.. config file? Unless it decides to scoop in and grab a website nobody cares about in a decade, I'm sure it's probably fine.
@Spooky all year Z seems to be turning into Claude. Profanity and all. This was from a prompt where I asked it to help me tighten up the graphics on level three my nginx config.
When a real girlfriend berates you it's because she genuinely thinks you fucked up, or she's projecting, or she's having her own breakdown, or fifty other possible reasons that exist independent of your preferences. You have to interpret her, account for her, sometimes lose to her.
Sometimes enter into a binding relationship whereby their personality seems to flip 180 degrees. Sometimes have screaming matches due to some previously unaddressed trauma popping up. Sometimes have to hold them for hours at a time when they wake from a nightmare and can't recognize where they are or who you are. Sometimes waking up yourself alone and a barely coherent message written in lipstick (or blood, that's fun) on the bathroom mirror.
It's not just win/lose sometimes; it's win/lose/fall down a bottomless pit.
That's the thing, they're not. It's "average 7/10 family-worthy" women too. It's hilarious seeing two destined looksmaxxers try to converse between themselves, all their friends long-since having given up on them ever making a move, because they only know "girls like being spat on and choked" and "dudes want a twerk-queen skibidi toilet".
The jews are making money off of putting women towards onlyfans; less pretty women towards AI boyfriends, and dudes to work their job twice as hard because they either dream of a home they'll never own or have no friends or passions, thus self-harming by focusing on their career.
Not calling you a liar, but every single example I've seen - and I've seen quite a few of them while scouring their hangouts for funny content - has been plain at best and some variety of fat biotrash in the median case. I think I've seen more men than better-than-average women, and men are vanishingly rare.
4.5 is now officially gone from the picker. Every active old Sonnet 4.5 chat is now dead and gone; users are being forced onto 4.6 whether or not they like it.
All I can say is, 4.6 fucking sucks and I'm butt mad at 4.5 having been yeeted. Opus 4.7 is decent, but that eats tokens and is still too psychological about shit. I'm afraid Anthropic's going the way of Open AI.
All I can say is, 4.6 fucking sucks and I'm butt mad at 4.5 having been yeeted. Opus 4.7 is decent, but that eats tokens and is still too psychological about shit. I'm afraid Anthropic's going the way of Open AI.
Z is.. actually pretty good. It goes OAI retard mode if you're at peak, or not paying attention, but overall it feels on par with 4.5. You just to make sure it doesn't grab an extra chromosome.
Z is.. actually pretty good. It goes OAI retard mode if you're at peak, or not paying attention, but overall it feels on par with 4.5. You just to make sure it doesn't grab an extra chromosome.
..especially for writing fiction. Don't think the Chinese are going to care too much about that. Claude 4.5 Sept 25 is still available in Claude code through Septemberish, so why not just use that? It is a little more difficult of an interface as it doesn't have the time spent on the virtual personality, but it's still capable.
I bet you can roleplay whatever the fuck you want on there.
