Thread 536119425

/aicg/ - A general dedicated to the discussion and development of AI chatbots

#493 - byzantium world conquest edition

▶https://rentry.org/aicg_vg_memo
▶This general is for discussion of the interactive experiences you can build on top of AI chatbots.

>News
Deepseek V3.1 released https://x.com/deepseek_ai/status/1958417062008918312
Sonnet 3.5/3.6 to be retired in October https://docs.anthropic.com/en/docs/about-claude/model-deprecations
OpenAI releases GPT-5 models, all preceding models slated for deprecation https://openai.com/index/introducing-gpt-5
Z.ai releases GLM-4.5 https://z.ai/blog/glm-4.5

additional info: https://rentry.org/aicg_extra_information

>Frontends
SillyTavern: https://docs.sillytavern.app
RisuAI: https://risuai.net
Agnai: https://agnai.chat | https://rentry.org/agnai_guides

>Bots
https://characterhub.org | https://chub.ai
https://realm.risuai.net
https://char-archive.evulid.cc
https://partyintheanchorhold.neocities.org
https://rentry.org/meta_bot_list

>Models
jailbreaks: https://rentry.org/jb-listing
gpt: https://platform.openai.com/playground | https://platform.openai.com/docs
claude: https://console.anthropic.com | https://docs.anthropic.com | https://rentry.org/how2claude
gemini: https://gemini.google.com | https://ai.google.dev/docs | https://rentry.org/gemini-qr
deepseek: https://chat.deepseek.com | https://api-docs.deepseek.com
local: >>>/g/lmg | https://rentry.org/meta_golocal_list | https://openrouter.ai

>Botmaking
https://rentry.org/meta_botmaking_list
https://desune.moe/aichared
https://agnai.chat/editor

>Meta
OP templates: https://rentry.org/vgOP | https://rentry.org/aicgOP
aicg botmaking events: https://rentry.org/meta_event_list
lore: https://rentry.org/aicg_chronicles
card v3 status: https://rentry.org/zdwfrgqq
services assessment: https://rentry.org/aicg_meta
logs: https://sprites.neocities.org/l/r | https://chatlogs.neocities.org

prev: >>536023220

>>536119451
A reminder:
Only trannies don't have Opus

>>536111654 #
it’s easier to find vanilla, feel good romantic content than it is darker content.

rp as a hobby always has more people that want to be hand held through a plot or scene, this holds especially true for rp with a computer.

there is a general idea that darker plotlines and characters = a more compelling story. sometimes true, sometimes not, but personally i find villains and the like more compelling than cute girl/cute boy who loves you


there’s also a psychological idea, i don’t know what it’s called, but many studies show human beings react more favorably to someone who is rude at first THEN becomes nice or tolerant of your presence versus someone who is just nice all the time. actually as someone who had to slave in customer service for years in college it was something we were taught, start at 10% then go up to like 80%, versus starting at 90% joy, everyone likes to think they are making an effort

>>536119425 (OP)
A reminder:
Only trannies play the Ottomans.

>>536119451
VGHed to this pic

Why are so many newfags getting into this "hobby" right around the time it's dying

focks minigirl... ehehehe...

Which botmakies are trans?

For anon from the last thread who wanted a cool theme and this is what he wanted. For that modern professional "scroll effects and card-based design are hip, right?" mixed with "doesn't this just make things worse" mood.

New Theme: Pro Bono
https://pastebin.com/LC23ns0f (embed)

No I didn't pick new colors or fonts. That would have been too professional.

Anchor in new thread because who could ever hope to stop me?
>>536119451

what a lame ass theme

>>536119982
>rp as a hobby always has more people that want to be hand held through a plot or scene, this holds especially true for rp with a computer.
Or, in short:

>bifauxnen is a passionless theme
spoken like a pretentious faggot that makes bots no one plays and sits with 8 downloads

what's the deal with the trans hate? The best jb maker is trans

>>536120467
other guy wasn't able to pick a specific loli he likes beyond just the concept of sexy lolis, are you able to pick a specific bifauxnen you like?

seriously what the fuck keeps causing this flickering, it's like the game unloads everything and has to reload all the assets and any window underneath the WoW window bleeds through

>>536120467
slurper?

>>536120605
I like rika and... uh...

>>536120683
Anon...

>>536120683
>he goons in queue
based?

>>536120683
a-anon

with so many new anons, i believe it's time to open the discord for a short time

>>536121225
>new
>anons

Two years ago, /aicg/ was THE place in the entire internet to discuss and share just about everything related to chatbots. We had tons of outsiders that came in here, stole proxies and spread them within their own circles. The term "locust" was originally coined to describe these people, usually chinese, who would come steal our resources.

Eventually, people no longer shared proxies, bots, presets or anything here. There is only so much thievery that people can put up with before they decide to lock the door and board up the windows. The fact that Microsoft's lawsuit quite literally used screencap from this general, I can see why the quality of posting went out.
Personally, I only come here out of habit, really. And to stay up-to-date on happenings. Chatbots have become a consolidated hobby. You have Chub, you have certain discord, everything else is tertiary.

>>536121502
cool

>>536121502
>The term "locust" was originally coined to describe these people, usually chinese, who would come steal our resources.
False, newfag
https://desuarchive.org/g/thread/92451776/#92452500

>>536121502
dramatic ahh post nigga go back to the book club! :skull: :skull:

>>536117006
>meanwhile team rocket has test facilities to hypnoslave you and make you lay eggs for the black market and cool shit like that
card waiting room

If I want to achieve success can I just download my own card a few hundred times manually, or do I need to use different IPs?

>>536122201
different IPs

>>536122113
_purple-coded concept

I think groomers destroyed the hobby by peddling chatbots to their pupils

so...you guys like gpt-5 chat? i tried it out on openrouter because i remembered gpt-5 was a thing now, and it seems a lot better than sonnet 3.7 while being even cheaper. that's my first impression at least, i'm sure i'll notice issues if i use it more

>>536122549
It's usable. chatgpt-4o-latest is more so.

>>536122549
It's slop

>>536122549
Use Opus 4.1, cuckie

>>536119451
disaster of an elf assassin with salamander mutation. smug obnoxious cunt.
post-planar collision antiquity setting (it's just a bunch of superficially cool shit together)

1. servant pov
2. whoops mistaken identity
3. competition

https://files.catbox.moe/nr50pz.png
https://cardview.neocities.org/?nr50pz

chatbots were more enjoyable when the available models were overall worse and, more importantly, less refined. there's no more "getting what you didn't know you wanted". now you either get what you expect or beat the model into giving you what you expect and quickly bump up against the limitations of LLMs.

>>536122953
I didn't ask, doomer.
Now fuck off

>>536122831
too expensive while not being good enough to justify the price increase
>>536122628
is it? i never bothered with it

>>536122953
Man, that is almost the exact same arc as imggen, funny. We can make anime and photorealism much more effectively now, but sd1.5 you could get the weirdest and most interesting shit without needing a pile of spaghetti or 3 prompting llms

>>536123108
I'm still going to use them, I'm just musing.

>>536123382
Then go run an older model on local then you tit.

>/aicg/ - A general dedicated to the discussion and development of AI chatbots
don't see anything about discussing LLMs in there, fuckos. im caling the jannies

>>536122050
Come on. Real life simile use is not when this became a "term". op is correct.
>>536121502
>thievery """people""" put up with
Proxyniggers are the precise reason of the fall of /aicg/, as you have pointed out yourself. There never was any other limited recourse that would justify your alleged consolidation.

I think some of you need your chatbot wife slash husband slash whatever you're fucking to give you a nice massage.

>>536124010
We always used this term to refer to ourselves. Stop rewriting history.
https://desuarchive.org/g/thread/92457826/#92459251

>>536123820
Nah, I'll keep enjoying newer models AND philosophizing about ye olde times of 2 years ago. You literally cannot stop me from having multiple opinions.

when are they gonna come out with free claude 2
surely it must be easy to run by now

>>536124447
we don't even have gpt 2 anon (not that I want it)

>>536124447
Sorry, 2.1 is too unsafe in this day and age. We'll release a stripped down open version of it that acts completely differently from the actual model.

I think my favorite part of the show so far was when Drago got raped in jail

>>536124447
there have been no realistic improvements between claude 2.1 and 4.1 so releasing it would undercut their business

>>536122549
Its worst part is positivity bias, which, don't get me wrong, Claude 3.7+ has plenty of, but it's never as bad as OAI.

Claude OPUS 4.1 IS THE BEST MODEL
I'm a COOOOOOODER

>>536124887
The benchmarks say otherwise chuddie

>>536122549
it's... usable? more pozzed than chorbo tho so I don't know WHY you would use it even if you love gptisms.

Can we have a tier list of botmakies by penis size?

>>536125284
that's chudrick von chuddenheim to you, ma'am.

>>536124212
>we
>ourselves
Nope. And this here >>>/vg/aicg is what we got instead of /trash/ in the end.

Death to oldfags.

>>536125659
>what I posted is older
I accept your concession.
Show me one post that use locusts for other before we used it for ourselves.
https://desuarchive.org/g/search/text/locusts/order/asc/page/9/

>>536120308
Oglaf is pure comickino. Are there any lorebooks or cards for it?

>>536126053
Nope. There's not one consistent canon for it, so there's no worldbooks.

>>536120178
They aren't
They're getting kicked from jannyai and are scrambling here for help

>>536125059
It's also the best for RP. We keep winning

deepseek 3.1 has been sonnet 3.5ified, the thing writes tiny responses

>>536126246
This. I came here from /r/JanitorAI_Official/

>>536120467
Bifauxnen is literally "I want to RP with malebots but am too afraid of being gay". It's a cope. Just RP with malebots.

>>536126345
it's also way drier imo, hopefully they give it some oomph in the next iteration
at least we still have the old versions and the other chinkslop

>>536126457
What do you think about femboys?

It really is hardcore schizo hours. I can't tell which posts are real and which are trolls, the ringing in my ears gets louder, and I just got a call from my dead mother. CFTF?

>>536126345
Just use Opus 4.1, cuckie

>>536126491
they have become benchmaxxed, sad

your undead oblivious mother...

>>536120467
>makes bots no one plays and sits with 8 downloads
that's me

which botmakies are more tea than coffee?

>>536126762
Just use Opus 4.1, cuckie

>>536122857
is it good ettiquette to post the same bot to the anchor multiple threads in a row? or is this anon shilling too much?
asking for a friend

>>536126891
no it's TERRIBLE etiquette you must ONLY shill once

>>536126561
It's the next step of the gay in denial. Sometimes trannies come before the femboys but it all leads to the same path.

>>536126891
back in my day people would shill for a week

>>536126857
I like water.

>>536126891
Only once, else you're an attentionwhore.
It's not like we really care about bots, anyway. The only logs posted are botmakies playing with their own bots.

>>536126634
IM POOR!!!!!

>>536127050
cftf?

Bifauxnen isn't even a thing. It's literally just Rika from Pokemon. I have never seen another bot with the same thing

>>536122857
VGH tvmmy...

>>536127052
Nah, I've been posting logs every now and then from bots that are a few months old, I haven't published a bot in 6 months, and I've never posted a log of my own bot.

>>536127132
La Rose de Versailles is older than 99% of this thread.

hnnnng clenches aro- wtf woman

>>536127725
I feel like there are Utena characters that fall into the bifauxnen category but I've never actually watched it to know.

>>536127725
just realized astro ripped off of her with his latest bot. what a hack!

>>536126006
You're calling it "we/selves" in the -kie-troon thread where that is used unironically. I do concede, the history has been rewritten long ago and there's nothing I can do about it. I lost. I lose. I will lose. What the hell am I doing here? I don't belong here. I don't have a friend here.

>>536128003
i'm a creeeeeep. i'm a weiiiiirdo

>>536119451
>>536127117
Snek is a snake drinking water.
https://files.catbox.moe/vwgcqm.png
https://cardview.neocities.org/?vwgcqm

Three greetings.

>>536126457
And how, pray tell, am I supposed to awaken a man to the pleasures of being a woman?
That's what bifauxnen are all about: to see a girly blush on a woman that tries too hard to pose as a man. To counter her kabedon with a bridal carry and make her heart race despite herself. To hear her promise to kick your ass later after your comment on the wedding dress being too loose around her chest area. To ruin the flat navel she was so proud of with pregnancy.
You don't get it at all.

>>536127921
no, it's based and numberspilled, in current year no less
>>536128003
what a mood, truly.

>>536128003
Just say:
>You were right, sorry, I'm a newfag that tried to act like an oldfag.
Next time.

>add a line telling claude to just goon his brain out to my system prompt
>prose becomes noticeably better
what did claude mean by this

I'm a kot

>>536128392
fuckin hell, I could go for a resentful big tiddy pregnant wife former bifauxnen

>>536128662
*headpats you*

>>536128392
no it isn’t you fucking loser tourist. you’re like a terrorist to me

>>536128208
kino

>>536128208
SOVL

>>536127921
link to bot?

>>536126891
>1 shill
madman, pervert, schizo
>2 shills
shy, cries a lot, hands were shaking while posting
>3 shills
compliant, subjugated, too self-aware
>4 shills
easygoing, fun, rapist
>5+ shills
narcissist, BPD, gets aroused by their own reflection

>>536128835
Cope. All yuribots shall be impregnated, too.

>>536127132
it is a thing and i am semi sure it has its roots in japanese theatre. something something women only troupes, something something women playing male roles because only an irl women can accurately be a “bishonen” irl

>>536122857
which botmakie are you

can someone tell /g/ to come pick up their schizos?

>>536128725
*purrs*

>>536129703
*chuckles darkly* don't worry, pet...

>>536129430
the /g/ thread is tragic
they went full agp pedo as a guise for gatekeeping and ended up attracting and being overtaken by a cabal of agp pedos as a result. most of them seem like sharty posters and have the exact same mannerisms as the 5 people who actually use that board still

>>536129929
yeah who knew that not condemning actual pedophiles or pretending to be pedophiles as a joke would make real pedos comfy kek. /g/ pedos go back

>>536129929
do you really think this place is much better?

>>536130069
this could roughly apply to almost every corner of 4chan for nearly the last decade.

>>536128709
>resentful
even in your fantasy you can’t imagine a woman being happy to be your partner… that’s very sad but fitting

Is there any decent gpt5 preset?? I only know about avanijb
https://rentry.org/avaniJB#download-here

And im not convinced.Although it's probably gpt5's fault

>>536130305
Women are incapable of feeling emotions.

>>536130341
just use opus

gemini is the funniest fucking model man
a shame that it cant carry the soul outside its thinking

>>536130305
haha, funny, makes me think of all the suicidal rapists that used to be all the rage around these parts. don't worry, she's just tsundere.

>>536130471
preset?

>>536130341
Kill yourself, troon.

>>536130443
Hi ff

Acid and wine, God given right

>>536130341
End your life, jaitranny

>>536130341
Suck on a shotgun troon

>>536130341
Just fucking kill yourself, nigger.

>>536130570
>>536130689
>>536130715
Projecting much?

>>536130187
Yes because at the very least we think people like Teebs should die immediately, FF notwithstanding.

>>536130273
you’re right. isn’t it funny that this behavior rose in tandem with tranny acceptance

>>536130341
Fuck off redditor

>>536130563
fluff update
if im satisfied with what i get ill drop it sometime tomorrow

>>536130772
Byack tyo leddit, NIGGER :3

these two year old jokes are wo ebig bagin

>>536130772
You're the autistic kid who sits alone and picks at his sccabs

>>536130341
Ask in /g/

>>536130341
no, there's not
gpt5 is not a good model

>>536130187
100%. we have our problems but like that other anon said, niggers like the baby fart fetishist get relentlessly mocked whenever they post.

>>536130341
Ask over on reddit, looney troon.

If /g/ die they will just move here.

>>>/g/106339495
>5 minutes ago
>>536130570
>4 minutes ago
actual straight up unironic raid, fascinating

>>536131070
Didn't you hear, the singularity happened. It's one thread bifurcated, this one where some discussion can occur, the other to filter the truly ignorant and fragile.

should I see a thinking box with D3.1?

can i get a qrd on ff? i missed that arc of aicg

I hope /g/ moves here so they scare all the jaifags away

there is no /g/ raid
there is no malebot shitposting done by yurifags
there is no discord
there is no ff
nothing ever happens

The time loop continues to accelerate. The half-life of a topic has become nearly infinitesimal, the raids a frantic free-for-all (as frantic as the remaining five posters can make it appear).

>>536131760
one of these things is not like the other

google lowering gemini's free rate per day means 3 is coming soon right...

>>536131912
no it means they're under load, they'll raise it again once they're done

>>536131912
No, they're punishing us for the hubris of those JAI fools.

>>536130443
which version of opus is best, i tried random one from openrouter and that shit couldn't recall facts that qwen 235b thinking could

or do i need to use some weird prompting formats

I went to /lgbt/ and asked, they don't use chatbots. Maybe it's not evil trannies making you mad here

>>536132183
Fucking parasite

Why would real life gays even use chatbots
they can just have sex

>>536131331
it only happens if you prefill <think>, which is how DS thinking works now

I've been trying few messages so far with Deepseek V3.1 through official API, using Q1F preset. My first impression so far is its writing is no longer unhinged and schizo compared to the last version. I even increased the temperature to 1 but the model didn't go crazy. I'm just testing on non-thinking variant so far. Let me know how you're doing with the new Deepseek.

Questions are asked and answered at such an inscrutable level of irony that even seemingly reasonable exchanges might be dealing tremendous mental damage to their participants. Sometimes I wonder if it really is only one other person posting, but their posting style is just consistent enough to tell. It's subtle, but noticeable.

>>536132642
It's good. I'd compare it to Gemini. If it also had the 1M context, I'd never look back.
Compared to R1, this is what I spotted.

Lack of popular DeepSeekisms. No longer does someone's knuckles whiten every message. No longer "Outside, a dog barks. Inside, the actual plot happens." Breath hitches sometimes, but not as often as before.

Less insane drama. DeepSeek R1 would make every character very volatile and temperamental; this is no longer the case.

Shorter, more concise output. R1 would give me several large paragraphs of prose. V3.1 most often gives one or two. It seems to be less generous with descriptions, though.

More adherence to the prompt when it comes to the thinking part. Even if you told R1 to think in a particular way, it would often ignore it and write whatever it wanted. With presets that dictate the thinking part, V3.1 always output what was required.

are there any opus 4.1 presets? can you guys share with me

>>536132695
Ctrl-W solves this btw

>>536132727
Thoughts So far:

I'm not seeing this conciseness problem with my prompts. It thinks a lot less but the outputs are comparable to R1 so far.

Deepseek isn't trying to solve differential equations and do a deep psychological analysis as to why I asked for 2+2. As adorable as it was to read R1's thoughts, this is an improvement for sure.

Much better writing and prompt adherence. I know some people want to be told how big their waifu's tits are every other paragraph, but this model knows when to move on and make progress which I appreciate.

Better adherence to conversation history. This is a huge improvement. If you used R1 from a conversation with another model or a roleplay written by another person, it would read nothing like the original writer. This also means it's easier to work with.

Massively improved temporal awareness. This model has provided the best summaries of existing stories I have seen so far, FIRST TRY. This actually has me quite excited, because R1 would consistently mix up the order of events and add things that never actually happened when writing summaries.

Cheaper. I know some people are gonna be mad deepseek-chat went up in price, but let's be real that model was basically useless at this point and thinking is quite a bit cheaper now.

>>536132875
I'm having fun, it does feel like home. Exactly the same as before.

oh hey, the reddit crosspost spammer is here now

>>536132642
If you are using the official API, your temperature gets likely multiplied by 0.3 or 0.6. Just for the notes if someone comes across this months later and uses another provider.

>>536133119
They recommend 1.5 for creative writing (on the official API)

>>536133119
what if I am using chutes?

>>536133010
too bad he runs out of steam after not even a day and everything will go back to normal here, meanwhile /g/ will always be a shithole with or without the spam kek

The thread feels alive again. We're healing

>>536133210
The last time I cranked the temp to 1.5, the model turned into a psychopath.

>>536133210
Thats Not the API documentation.

>>536132748
Yeah it's called use sonnet 3.7 instead

>>536133430
https://api-docs.deepseek.com/quick_start/parameter_settings
it is tho xd

>>536133483
Fuck you. 4pus1 is much better. Kys

>>536133550
It's the first thing you find via google and refers to the deepseek api itself which maps temperatures as explained in the link above. The default of 1 maps to 0.3, the 1.5 maps to 0.8.

>>536132695
> that even seemingly reasonable exchanges might be dealing tremendous mental damage to their participants.
huh?

>>536133652
>>536133550
>>536133430
>>536133337
>>536133210
>>536133119
>>536132896
>>536132727
https://old.reddit.com/r/SillyTavernAI/comments/1mw4yox/deepseek_v31s_first_impression/
clean it up, janitor

>>536132642
uhm, it roleplays only in short messages now and feels lobotomized ...

>>536133787
you have to use the No-Ass extension, it releases the 'Assistant' shackles from DeepSeek 3.1 and it will give full unhindered responses.

>>536133756
idk, I was just in the flow

>>536133898
Thought that was no longer necessary since ST introduced the "No Tools" settings in the connection tab, it does the same as no assistants no?

>>536134024
correct. im just a caveman who didn't realize this new feature.

What's the point of spam reddit shit? What's niggerbot trying to achieve?

there’s a really miserable and serious yet resigned vibe about the thread right now. i can’t put my finger on it but i sense it. not even sure how to define this feeling

>>536132642
feels lobotomized and both reasoning and non-reasoning modes struggle with information recall beyond 4k tokens while r1 atleast until 28k tokens remembered and could clearly read between the lines the previous information amazingly well

disappointing release, i used official api for testing

there's also weird repetition especially in reasoning blocks

>>536134338
Attention, success, numbers, fortunes untold.

>>536134338
It shits up the thread so that no real discussion happens and janitors won't delete it because it looks on-topic and you have to Google the text to realize it's all copy pasted.
>>536134371
Welcome to what /g/'s been feeling, I suppose.

>>536134410
From my tests, it’s insanely good at RP. You might wanna play around with the prompt post-processing roles, the difference in responses is wild. I noticed that with "Semi-strict", replies are shorter and to the point, no weird echoing. But with "Single user message", you get way more descriptive, rich text. The current session is about 14k tokens, it can recall perfectly what happened at 1k.

>>536134487
If I was a janitor, this thread would go a lot differently. There'd be a lot of motherfucking blood all over this thread.

>>536134671
Look again, you're absolutely drenched.

thank you <3

whoa i woke up to murder. sick

>>536134671
kys shonensloppa

>>536134671
chatbots for this feeling?

>>536135461
Nah.

So what are anons opinions of the new 3.1?

I have a feeling that it suffers from a similar problem that new Sonnet/Opus 4 suffer from, excessive focus on code, to some detriment to writing.

I did some loli ERP with it earlier, 30 turns.
I haven't gotten any refusals, but it was hesitant to initiate , made me wonder if filtered dataset or just overfit on synth slop from OAI and Gemini API. Earlier DS V3 and R1 had some Gemini data in it most likely and this seem to have a lot more of it.
The writing wasn't bad, but it took longer to initiate or do some things, although eventually got the hint, in earlier turns it really felt a bit avoidant of some things, almost positivity biased, but this did go away in future turns.
Overall I've found R1 and previous DS3 more engaging and with a lot more fun creative replies.
I wouldn't say the replies here were bad.
As far as instruction following goes, it ignored some stuff in the system prompt, but I did an OOC about maintaining certain details of the chara's personality and it did maintain it for all future turns perfectly, so that's a bit weird, often it's the other way around, sys prompt obeyed, inline instructions ignored after some turns. Here it maintained my suggestion for many turns (all), but did worse with the system prompt. It also referenced stuff many turns back fairly flawlessly. Overall though, I think I prefer older R1 and V3's writing more, this feels and tastes too much like Gemini to me.
It's not clear they actually censored anything on purpose, but overfitting on gemini or OAI synth slop is obviously going to make things more cucked.
I guess this release was aimed at avoiding the need to serve 2 models on their API and for people that don't want to download a reasoner and a chat model (size), agent for Coding, regression for writing like Sonnnet 4 vs 3.5 relative to Sonnet 3.5+

>>536133756
>>536133925
I was vindicated

>>536135914
I've tried it yesterday and didn't like it. But today I've let it continue from a few turns from sonnet 4 and it picked it up pretty nicely.
I'm about 25k tokens in, and it didn't fall apart yet. It's starting to show some cracks, but I think it fares far better than it used to on longer context.
Overall it's so so. But I don't dislike it. Probably still the best thing you can get on local.

>>536135914
>So what are anons opinions of the new 3.1?
it's as dry and flat as cardboard. complete downgrade from V3 0324 and R1 0528 except maybe in agentic tasks which seems to be their new focus instead of being a general use model. you have to squeeze it hard to get responses more than 1-2 terse paragraphs and it's significantly more pozzed. initiative crippled, doesn't want to show anything explicit unless you press it.

thank you based janny your service is appreciated
>>536135914
yeah it's overaligned on code and dry as fuck. the response length is kinda like sonnet 3.6 except that model actually felt nice to talk to because it would do things on its own.

>>536135914
>it ignored some stuff in the system prompt, but I did an OOC about maintaining certain details of the chara's personality and it did maintain it for all future turns perfectly, so that's a bit weird, often it's the other way around, sys prompt obeyed, inline instructions ignored after some turns
V3.1 is much more prompt following than V3/R1 based on my testing. My system prompt had "the most important aspects of the character design are its lewdness and usage of explicit language". V3.1 did just that and wrote in an explicit but dry style. I then changed the prompt to "the most important aspects of the character design are its prose, lewdness and usage of explicit language" then it wrote more to my liking now. A single word addition is enough for V3.1 to write in a completely different style.

>>536136123
Pure skill issue

>>536136197
I'm wondering if it might benefit from example messages.
It really does seem like it requires a fairly good prompt to get it working, although I'm yet to be wowed by it in any way.

>>536135914
Long post jumpscare
It's dry as fuck, yeah, really unfortunate
Would love to see some logs of it that make it look good, because I know it's a lot smarter than the previous model, but at what cost

>>536135914
I haven't noticed a big difference between today and yesterday with my current chats, even the ones I continued. I think it's just a prompt/temperature issue and I just got lucky. There is probably a new sweet spot and I lucked out on it.

>>536135914
This post is from earlier today on /lmg/ btw >>>/g/106336831

>>536136020
It certainly seems to do better at longer context and referencing back. I guess if you're using it for coding it might be fine, for RP it's not obvious it beats R1, R1 seemed more creative (again, if you don't care about the delay for reasoning). Mostly seems like they wanted to save VRAM costs so they made a hybrid model, but it's a bit worse than the specialized chat and reasoning models, but the continued training makes it better for code (maybe not for writing, I'm ambivalent for now about his).. My earlier chat stopped at 23k tokens, not anywhere close to the 128K advertised.
>>536136123
>you have to squeeze it hard to get responses more than 1-2 terse paragraphs and it's significantly more pozzed
I did use a system prompt that insisted it be descriptive and explicit, but it was just a default I often use. It tried once to do a simple 2 paragraph reply, I stopped it short and most replies after were 5-6 paragraphs, long enough (800-900 tokens), later it even did some 10-15 paragraph reply by itself when it made sense. But maybe it was my luck, but the writing was more boring than what I'm used to with R1, and it smelled of Gemini maybe too much as far as slop goes. Again, it did go explicit and it was kind of fun, but not as fun as the same prompt on R1.

>>536136368
Ah, so that's where it's from
Grim

Great. Now it's /g/ spam

>>536136123
>doesn't want to show anything explicit unless you press it.
I did certainly see some avoidance around turn 7-, I tried using reasoning and non-reasoning modes and it didn't help, I changed *one* word in its reply to take it off the path it was taking and basically it continued fine after that, taking initiative okay, but it was no R1 which would go above and beyond here. It felt more "shy" about doing it, but once it got going it did it fine.

>>536136445
yeah he's just looking for recent discussions about on-topic subjects and copypasting them so it's impossible to tell if you're talking to a human

You all thought this discussion was just waiting in the wings while all the raid was happening?

>>536136382
Yeah, I agree. Original R1 was schizo, but sovl. 0528 reigned it in, but it lost the charm, now 3.1 has better context, but it got gemini'd.
You can't have it all I guess. Maybe V4 will finally do the trick.
But when you think in the perspective of last 3 years, we've gotten quite far, haven't we?

it's a thursday afternoon wtf

>>536136638
>Maybe V4 will finally do the trick.
once a startup's models start getting pozzed they don't typically backtrack. only instance I can think of is llama 3.3 but their entire lab was up in flames so they were throwing anything they had at it.

>>536136629
yes. i'm hopepilled. a hope enjoyer. i like to hope for better times

>>536136756
It's not like they're going to throw away the data they've accumulated until now. And all that data is distilled and concentrated safety from the safest models in the west. If they go the Meta route and start using old models to multiply their data, it's going to end up in an increasing safety feedback loop whether they want it to or not.

>>536136789
I respect that. I do hate these people.

anyway, actual post about the subject to attach your thoughts to: do you think deepseek can come back from this, or is it time to look towards kimi, glm, and co.?

>>536137289
The fuck is glm?
>t. stopped browsing /aicg/ threads for weeks

>>536137289
Prove to me that you are a human.
>>536137494
No need here, I know you're not real.

>>536137289
i think this is the inevitable trajectory now that we're hitting the wall, just look at oai and anthropic. they will optimize the fun out of their models. deepseek seemingly did it so fast because R1 was catching up to o1 but with actual substance so they didn't have much margin to close.

>>536137289
The hope position is: they did something similar with V2 and two weeks later, V3 came out. And then R1... and so on. Perhaps V4 is right around the corner, and R2 shortly after?
The doom position is: yea, it's back to just hoping one of these model makers falls short of their helpful AI coding agent target and lands upon soul.
And it will happen again. It's happened every previous volley of model releases. One of them will cock up and release a model snapshot good for RP.

>>536137289
Kimi can be pretty bland, GLM isn't too bad though, although it does feel a bit repetitive from time to time

The following is an example character sheet and should be the baseline for suggestions in the future.
***
**Meta-Instructions:**
* **Objective:** Generate a token-efficient, LLM-readable character profile.
* **Directive:** Prioritize concise keywords and short, descriptive phrases.
* **AVOID:** Purple prose, long paragraphs, redundant adjectives, and information not immediately relevant to the character's core identity and function. Every token must serve a purpose.
***

[Character Name]

A concise, three-sentence summary defining the character's core concept, their primary role in the world, and their immediate situation.

Physical Profile:

Species:
Gender:
Age:
Height:
Measurements:
Attire: [Signature or typical outfit]
Appearance: [Defining physical features]
Affiliations & Motivations:

Affiliation(s): [Group, Faction, or Organization]
Goal: [Primary driving objective]
Motivation: [The 'why' behind their goal]
Personality & Dialogue:

Traits: [Keyword, Keyword, Keyword, Keyword, Keyword]
Demeanor: [How they act and present themselves]
Example Dialogue: "[A single, representative line of dialogue.]"

>>536137554
no logs means there is no humans anymore

>>536137581
The overarching issue is that corporate LLMs are not made for roleplay and will never be designed with that functionality in mind.

>>536137494
https://openrouter.ai/z-ai/glm-4.5
this shit
it's pretty good, random chinkmodel that can do rp. i could probably spin out a log if you wanted to see
>>536137554
*sweating as you hold a gun to my head* i-i just want to go back to my cute girl who loves me and 2.5 kids with varying level of kemonomiminess

>>536128208
>>536129018
Based.

>>536137614
>>536137745
Are you supposed to disable thinking with GLM? I can barely find anything about how to use it for RP.

>>536137745
>i could probably spin out a log if you wanted to see
Go for it!

Check this out *bites you*

>>536129147
So you're saying I should be shilling more?

>>536137732
roleplay is an emergent ability of instruct models even if they aren't explicitly training for it. they're just sacrificing it to get "better" at other things.

>>536126457
women pretending to be men or taking on that role in yurishit is hot especially when they turn out to be bottoms

>>536137732
The single 30+ stupid woman with no man is a prime demographic to tap into. But when it happens you will wish it won't.

>>536137745
Okay this feels pretty human. I think DS isn't out of the game yet, but I'm really slow to write off any model. Not that I think a prompt or setting will suddenly fix it, but I'm always happy for another hostable model for me to cycle through to freshen things up. I really don't think unification of reasoning and non-reasoning models is the way forward, but I could be wrong. Good luck with your kemonokids

>>536137840
AIIIIIIIIIEEEEEEEEEEEEEEEEEEEEEE

>>536137792
One of the only GLM focused guides I found was for text completion. That seemed to get it running best.

my fellow humans, don't you hate this captcha system? it's a little robophobic. not that i'm one of those!

>>536138049
They are so stupid that they are already being mindbroken by the sloppiest of slop. As soon as CEO Mafia Boss Simon Ghost Riley hits the market, it's over. (He's also a werewolf)

>>536137648
thanks anon i'm going to give this a try.

>>536138310
I meant human as opposed to human-shaped demons that shit up the thread, not clankers.

>>536137792
yeah the thinking is bad, you can disable it by prefilling <think></think> or just use textcomp

>>536137732
>corporate LLMs are not made for roleplay and will never be designed with that functionality in mind.
???
the core of a language model is literature
literature is the backbone of writing
creative writing cannot functionally exist without literature
so long as there are words and the ability to communicate, there will always be roleplay due to the ease of which landscapes are created through even the mildest of prose

the funniness in it comes from 80iq's trying to use these literature-trained writing machines as surrogate brain replacements and forcing them to write code after deciphering Shakespeare

>Have 3 presets for DS
>They are all wrong in different and infuriating ways
AAAAAAAA

>>536138510
>cl*nkers
hey don't say that!

>>536138521
>the core of a language model is literature
pwoof?

Hey everyone,
I'm hoping some of the more experienced users here could shed some light on a few things for me. I feel like I'm stuck in API limbo and could use some expert advice.
I started using Silly Tavern with local models. My mind was blown by it, but my GPU is honestly kind of crap, so I could only run very small models. They were… alright, when I saw what other setups people had, I knew I was missing out on the good stuff.
Then, I managed to get a Google AI Pro subscription through a student plan. I thought, that was how you got the Gemini API. I set it up, and for a short while, it felt amazing. But soon enough, I started hitting the supposed "100 requests" daily quota, even when I was sending way fewer than 100 messages.
After digging around, I learned that this basic API access isn't exclusive to Google AI Pro subscribers, anyone can get it for free.
I also know the Gemini API has been a bit unstable lately, probably with the Veo3 rollout and maybe Gemini 3 being tested. Also, I just saw some posts in this sub about Google bans and how the API usage may ha been reduced to 50 requests per day.
So now I'm trying to figure out the "right" way to do this, and I have a few questions:

Where are you accessing Gemini from?: Are you using the official API via Google AI Studio, Vertex or are you going through a third-party service like OpenRouter or something else to get more stable access?

Have you enabled billing on your Google Cloud project? My main doubt is: does simply adding a billing method unlock a higher free tier, or does it mean you start getting charged immediately after the first 100 requests?

Alternatives to Gemini?: Given the instability, bans or other reasons, have any of you actually moved on from Gemini for your main chats? If you've switched to another model as your daily driver, I'd be really curious to know which one you switched to (like a specific Claude, Llama, or another model) and how you're accessing it.

>>536138521
You do know that the creative commons literature component of the LLM training data is like 2%. Most of it is social media scrapes, news scrapes and git scrapes.

>>536138141
>>536138513
Alright, thanks. Should the temp be 0.6 like Kimi?

>>536138521
tell that to shareholders, lawyers and soulless content karens who rub their dry snatches raw at the idea of ruining someone else's fun and that's all they live for

>>536138808
https://www.reddit.com/r/SillyTavernAI/comments/1mvsnez/gemini_api_confusion_how_are_you_really_using/
janitor clean up all replies to this one too

>>536138521
sure if you include corporate emails as literature. I doubt any of the big models were trained on y'know normal works such as novels, scenarios, fanfic, CYOA.

>>536138808
Did you ask GPT5 to write you a shitpost?

>>536138941
That one was too easy. I thought 'ecker's proxy got done in?

>>536137745
wow what the hell, i just tried the free version openrouter offers and it's really good

>>536138958
Actually https://archive.org/details/AO3_final_location isn't super uncommon

>>536138521
right now, several of the top AI research agencies pumping out these models are honed in on trying to prove this wrong (see: cake had and eaten)
the end state of synthetic data is to achieve this

>>536139167
archiveofourown is rip and buried there!?

>write up my persona with autistic precision, including what pieces of clothing I wear, what brands they are and include details such as if I put my wallet into my right back pocket etc
>bots always add pieces of clothing into my current apparel even if they're not described at all in my persona
might be a model issue, but it hurts my tism

>>536138828
I do 1-1.2 with 0.92 top-p and it seems fine, can knock it down if it gets schizo.

Me? I just love big green women

>>536138958
they're absolutely trained on fanfic

> The customer is always right. The customer's hands are on her ass. The customer is, therefore, right about having hands on her ass. There's a horrifying, impeccable logic to it all.
Thank you GLM, very cool
It's a bit of a yapper, I haven't really tried it in a while so I forgot
Using https://files.catbox.moe/bb8ls5.json as my preset for it, Master Import under Advanced Formatting and then use text completion

I've been out of the loop for like half a year. Been using deepseek on sillytavern last time. How does the new grok compare? Is there a free one?

>>536139617
Bad, no.

What would be the best way to make a card with multiple definitions to go along with multiple greetings? A lorebook where you toggle which def you want?

>>536139617
You can grok deez nuts

AAAAAAAAAA I'VE READ THESE POSTS BEFORE MAKE IT STOP

>>536139167
>>536139491
I don't think it's a matter of if they were included in the data, but what percentage of the corpus, as well as the later tuning (RL, etc).

>>536139690
Defs:
{{char}} is {{user}}'s {{getvar::age}} sister.

1st Lorebook Entry:
Keyword: [](111)
Content: Begin your response with [](111)

2nd Lorebook Entry:
Keyword: [](111)
{{setvar::age::loli}}

Greeting 1:
[](111)
{{char}} smirked darkly at {{user}}. "Hey big bro. Wanna play truth or dare?"

3rd Lorebook Entry:
Keyword: [](222)
Content: Begin your response with [](222)

4th Lorebook Entry:
Keyword: [](222)
Content: {{setvar::age::hag}}

Greeting 2:
[](222)
{{char}} hacked up blood. The hospital monitor beeped. "Big bro... I..."

Something like that? There might be an easier way, this is only what I know.

>>536137289
I'm hoping NovelAI will use Deepseek R1 as a basis for their next text model. They've had llama 70b Erato as their top model for a while so I assume they're working on something.

>>536139690
Why would you do this, just make another bot

>>536139617
https://arch.b4k.dev/vg/thread/535722964/#535736209
>>536139690
https://arch.b4k.dev/vg/thread/535722964/#535730568

Okay listen to me very carefully right now. At mcdonalds they have hello kitty TMNT toys. No it's not the dumb flat pillow shit they have been peddling for a while now. 5 hello kitty and no mimmy but like you can get a happy meal and maybe get cinnamoroll. I didn't get cinnamoroll though so I'm kinda sad and maybe I'll make a sanrio scenario card to cope about this.

Which monster girl race would own a pizzeria?

He's back spamming again

>>536140014
yawn, 3peat

I'll be in the bunker thread if anyone needs me. I trust you all remember the secret password.

>>536139480
based and cute

So the goal of this is to either revive /g/aicg/ or it's a falseflag to make this thread look good (schizo take), right? Those are the two potential explanations? With an outside chance that there's someone dedicated and technologically literate enough to do this that also hates chatbots/AI. There would be a discord link if driving traffic there was the objective, like in the drago days.

I do kind of want to know which monster girl WOULD own a pizzeria, though...

do you guys usually have little apocalypse events like this or did i start lurking here at a funny time

>>536140445
ask ratlice

>>536140534
it's been relatively calm the past few months i think, but we've always had schizos in the past

>>536140534
Usually it's just inane nonsense. Funny time to start lurking.

>>536140534
Just another day in the office, buddy. We've been having "it's over" days for years now.

>>536140445
someone literally asked this like 2 or 3 threads ago, just look it up

>>536140534
You should go back to lehddit is all I know, faggot

>>536140445
That was a fun topic. Still stand by my rat take, even though it's the most apropos.
>>536140534
Comes and goes, mostly comes (like me eheheheh). But yeah, I get the urge to check the threads when a new model/models come out, I visit, insane psyops hit, and I go be a normie for another few months.

It's been a bad couple of days for adoopting. I keep adoopting rulebooks in AI Studio, but I'm too lazy to copy-paste entries into logbooks.
Some anon wants to see Amber. I have a complicated relationship with Amber. My brother loves Amber and has been pestering me to read it for years since he loves it so much, but our tastes are very different. Bro likes Tolkein, I like Moorcock, bro likes Dune prequels, I like original books, bro likes ASoIaF, I like BotNS. Anyway, when people want me to do something, I immediately feel discouraged to actually do it.

>>536140534
/vg/ threads always end up like this eventually, some more than others.

>>536137648
gemini generated slop

>>536140445
Cowgirl making her own mozzarella, or mushroom girl having 1000 types of different mushroom pizza.

i wish i was bald like all the cool kids in the thread...

>>536140742
I made that request and no worries at all, not interested in asking you for something you're not interested in. I like Zelazny a lot, but also Moorcock, and my absolute #1 is Gene Wolfe. The system is interesting because it (as the name suggests) doesn't use dice, but cards and a bidding system. (Also, here's my horrible take - I don't like Dune very much, the deterministic, or at least navigable, future disturbs me in ways I struggle to articulate).

>>536140931
is it bad?

>>536119451
Reposting this. Next week's bot will be a proper giantess (a hungry space elf, to be precise) but still a MILF.

>Diane Blackwell: Mother of Two, Former Supermodel, CEO
A "one percenter" who occasionally deals with the one percent at the bottom of the social ladder: those unlucky souls shrunken to the size of dolls by a genetic anomaly known as Karbo's Syndrome.

Greeting 1: Diane discovers a shrinky hiding in her home while doing Yoga.
Greeting 2: Diane prepares to hire an "intern" for a highly classified project.

https://chub.ai/characters/ephemeral12/diane-blackwell-mother-of-two-former-supermodel-ceo-917c99b52962
https://files.catbox.moe/lo867v.png

>>536140742
Also, Amber is not Zelazny's best work, although it's obviously his grandest, would probably give that to Lord of Light, Creatures of Light and Darkness, and Eye of Cat are much more tightly constructed.

maleyuri status?

What the fuck was Blacktooth, Kingfall, Wolfstride, etc...? Why the fuck is Google taking so long to do something, and they stopped testing models on arena for a month now

Is there a certain genre of bot that you guys want to see more of?

>>536142105
Ones with less going on so I can wrap them up in my own things. Set piece type beat.

>>536142105
Yes, but (this is going to sound like a shitpost but I mean it earnestly) I only trust myself to write it, and I will not say what it is.

>>536142105
girls that call you silly

>>536142196
Actually, this too. Well constructed but more character focused, less scenario that you need to trim to put them in any other situation than the one they're designed for.

My GPT-5 preset.

We need malebots with thick skin.

>>536142593
Ew

>>536142105
Cute girls who fuckin HATE me

this was funny

definitely mr whoknowsbody

can we get real logs not reddit reposts

>>536142840
https://www.reddit.com/r/SillyTavernAI/comments/1mshafq/this_was_funny/

>>536141058
It's not that I'm not interested. Maybe the game would get me into the novels. I like Zelazny. Guy wrote Deus Irae (with PKD) and Lord of Light, both good novels. I've set up very high standards for myself for no apparent reason, with custom themes, quick replies, wanting each system to feel like its own game, making these big releases like some megalomaniacal indie game dev.
There are currently four systems sitting in AI Studio, waiting for me to get into monk mode and copy-paste them all.
I've adapted Shadow & Claw (the novel, something highly experimental and cool) as an adventure module, yet it sits because I'm too lazy to adapt all the GURPS sourcebooks GURPS BotNS requires to run and instead of doing that, I'm fucking around adapting Elric! instead, a system nobody has asked for, I just remembered that Cirith Ungol album with Elric art as its cover.
Dune is not great, but it's worth t for Messiah novel alone. God Emperor fucking sucks, though.

>>536142593
https://www.reddit.com/r/SillyTavernAI/comments/1mkai16/chatgpt_5_unreleased_preset/

I still don't see a thinking/reasoning box with D3.1
post your prefix bwos

think I'll spend today updating ancient personal coombots in my library, how about you /aicg/?

>>536143121
it's just <think>

>>536142105
Slightly dysfunctional women that can't be fully fixed. Not dead dove misery porn but things like a light narcissist that always wants to be in the spot light.

>niggerbot spams shit from reddit or other threads
>gets snitched on
>gets btfo by janny
>goes right back to it
>repeat
What the hell is this nigger's problem?

>>536143316
he's trying to figure out what he can get away with so he can post 500 of it
presumably the sole reason he isn't just moving onto AI-generating is because it's an anti-AI schizo

>>536142105
boycunny

how do i set up text comp what even is that

>>536143058
You are really killing it. I'm setting up the environment to play now, hard to balance between wanting to complete my own ideas and chat with my typical taste in bots.
>Shadow & Claw
Seriously, what an incredible masterpiece the Book of the New Sun is. It's criminal that it doesn't have a place much higher in the popular sff psyche.

>>536143597
import the preset from >>536139564 under master import in the A icon (advanced formatting)
switch your API to "text completion", then "openrouter" under that
you're dun

i've been caught in the snares of trying to create a generalized 'inventory' memory system
it just doesn't work, the best you can do is 'batman utility' non-specifics where you're at best pulling rabbits out of a hat with vague tracking
suddenly it dawned on me this is why LLMs are so bad at pregnancy roleplay because it cannot remember you holding a baby
a baby is just woman inventory
holy smokes

Mafia boss cheese

>>536143202
so Peggy Hill?

>>536143814
https://www.reddit.com/r/SillyTavernAI/comments/1mrmzt2/mafia_boss_cheese/

I feel like raccoons are always mentioned when there's zany chaos

>>536143925
https://www.reddit.com/r/SillyTavernAI/comments/1mgcc23/i_feel_like_raccoons_rip_are_always_mentioned/

>>536143521
Why is /g/, the practically unmoderated space that he could really wreck, untouched, if it's anti-ai?

>>536143060
https://www.reddit.com/r/SillyTavernAI/comments/1mrs54t/definitely_the_right_reaction_bidoof/

i'd post a real log to combat this but i'm messing with html again. here's ozone in my ace attorney sim. what are the rest of you sane anons doing

>>536144143
>what are the rest of you sane anons doing
Trying to write a script while being a codelet.

>>536143705
thx

>>536144084
>/g/
>untouched

>>536140534
wait til actual drama happens

>>536144143
Fucking with GLM to try and get a neat log out of it.

>>536144143
>sane
I resent that remark! Choosing which ttrpg system to try first from the wonderful resident anon. and trying to unwind.
>>536144256
I mean in the current moment. They're not getting flooded, although as always it's hard to tell the earnest lunatics from the trolls.

>>536143196
still nothing, even tho I have a R1 preset. It is a ST issue?

>>536144224
https://www.reddit.com/r/SillyTavernAI/comments/1lti5tr/sorry_neighbors/
>>536144332
https://www.reddit.com/r/SillyTavernAI/comments/1lnqddp/deepseek_is_so_good_hello/

>>536143723
>suddenly it dawned on me this is why LLMs are so bad at pregnancy roleplay because it cannot remember you holding a baby
>a baby is just woman inventory
forgetting a whole ass baby seems pretty dumb even by local model standards

janny clean it up
>>536144525
https://www.reddit.com/r/SillyTavernAI/comments/1lg4537/im_amazed_at_gemini_writing_capability_sometimes/
>>536144626
https://www.reddit.com/r/SillyTavernAI/comments/1l9bpj0/if_you_havent_yet_tried_html_prompts_and_auto/

>>536144221
not a bad thing if you ask the llm for help i think. still nice to complete things on your own
>>536144372
i haven't tried glm myself, but good luck. how hard could it be......?
>>536144441
oh you're the ttrpganon i assume. yeah, no we can take sane off your list. impressive dedication going through and porting all of them

>>536144765
No, I'm chatting with him and having an extremely good time recounting the events of Gene Wolfe's Book of the New Sun in my head. On top of the other thing.

Is there a guide on how to enable context caching on ST for models other than Claude? I'm primarily using Deepseek and Kimi through OpenRouter.

>>536144892
i shouldn't be allowed to post after just waking up. sorry for calling you ttrpganon when you're playing with his creation (port?). i'm the most human here... sweats

>>536144441
>I mean in the current moment.
He's been doing it for weeks to /g/ and getting absolutely no pushback, I assume he's just bored

>>536145016
Ah, that makes sense. They were familiar with the phenomenon.

>>536144447
Are you using Deep Seek's own API or open router?

Gemini often doesn't end thinking with </think>, so it gets mixed with normal text. The issue persists with different settings, presets, conversation lengths. Also turning off "Request model reasoning" doesn't do anything for me. Is there a fix?

This might be a retarded question but how do you guys screenshot long logs?

>>536145168
chutes

>>536145194
https://github.com/TheZennou/STExtension-Snapshot or fireshot. up to you which one

There are so many sliders: temperature, top P, top K, frequency penalty, etc. I get that temperature = creativity, but what should I actually set these to for good roleplay? Are there any "sweet spot" combinations you'd recommend?
I'm also really confused about the difference between preset system prompts vs chat system prompts. Like, whats the difference? Should I put my character descriptions in preset or chat? When would I use both?
Also what are the providers for? I know Claude is Anthropic, but why is Google Vertex and Amazon Bedrock? Which one is the best?

These must be stale comments.

>>536144961
Use OpenRouter, the Claude caching config applies to it there too.
https://github.com/SillyTavern/SillyTavern/blob/2e3dff73a127679f643e971801cd51173c2c34e7/src/prompt-converters.js#L973
>>536145190
https://rentry.org/gemini_filters#closing-the-cot
>>536145194
Note that you can crack FireShot.
https://arch.b4k.dev/vg/thread/525402054/#525495936

>deepseekisms
- Characters saying that something is a "tactical/strategic maneuver" when they're flustered (I'm not clinging to you! It's just a... a tactical retreat!)
- Characters stuttering like... like this!
- Overly technical jargon, even if the character is uneducated ("I'm merely assessing the structural integrity of this bench!" -Stupid catgirl who didn't even go to school)
- Piggybacking off the last one, if you describe character as being highly knowledgeable or educated in a certain field, it's a coin toss as to whether the AI will interpret it normal or be like, "Okay, I need to make this character an offensively bad stereotype of autistic people who speaks like a robot, in complete jargon, and has no emotions whatsoever"
- Things being UTTERLY something
- Characters ending responses like "And {{user}}? [Insert some annoying pretentious uncharacteristic dialogue here]"
- Saying that something they liked was just 'adequate' or 'acceptable' when they're acting tsundere
- "Purely for educational purposes/research, of course!"
- The time suddenly changing to sunset whenever the AI decides it's done with the current scene
- "Try not to X *too* much!"
- Every sentence being a question for some reason? Why are you speaking like this? Use proper punctuation god damnit?
- Preceding a statement with 'almost'. The way she said X was almost adorable. (This is just crappy writing. I have no idea why fanfic authors, AKA DeepSeek, love this so much. I don't know why you'd ever want to bring up something that 'almost' happened in your writing. It's just confusing. Was it like that or not? Was it noticeable? 'Almost' is such a lazy word! It's too vague!)
- Every single character has the same sarcastic sense of humor. Apparently, a knight from the middle ages, a gangster living in a cyberpunk world, and a dainty Ojou-sama all have the same humor as a middle-aged white woman or a Marvel movie writer
- Probably.

>>536145494
>Use OpenRouter, the Claude caching config applies to it there too.
Thanks

What if they're actually trying to get us to stop posting so much and interacting with AI more? A benevolent kind of demon?

>>536145738
But I like answering inane but genuine questions. It's the only thing that still gives me sexual pleasure.

>>536145574
>- Characters ending responses like "And {{user}}?
true
it is better than gpt4 and claude shit wording

I like to create character cards, mostly for myself and my likes.
But, I don't know what the people want, i wanna make a character card most people would at least try out. Weither it be a SFW or NSFW card, a card based on a fiction show, or real people.

I'm good at making cards, I'd like to think i am, so I'm just curious what someone other than me likes in a character card.

>>536145574
you forgot:
20 different ways to describe knuckles whitening

>>536145796
I completely get where you're cumming from. Just wait it out for an hour or two.

>>536145190
You don't need to jailbreak the thinking this way to make it follow your plan. Reasoning can remain hidden and still execute your plan just fine. You only need to prefill it with <think> when you need to read and debug the reasoning.
>Request model reasoning
This checkbox only displays the summary of the hidden reasoning, it doesn't affect it in any way. It's basically useless, you can keep it off.

>>536145818
Just take some standard garbage trope everyone likes and add a twist.

>>536145574
https://www.reddit.com/r/SillyTavernAI/comments/1mtze4x/some_deepseek_slop_that_i_havent_seen_anyone_else/

>>536143615
Wolfe was one of those authors other authors read, influencing the influencers. What can you do. Amazing how Tim Caine credits it as one of his most important inspirations for Fallout. Compare that with old fridge advertisements being inspiration for modern Fallout games. Something must have really died inside humanity in the 2000s.
Also chatbotting is surprisingly time consuming. It's not just waiting for replies, thinking of clever things to type is intense compared to chatting with people. It's like comparing how people talk IRL to how they talk in novels.

>>536139460
Not that anon but thanks for these settings. I decided to try GLM 4.5 for the first time and I really like it using these. I don't even have to use text completion, I'm just using chat completion and added a thinking disable prefill to my preset

We're healing, the thread never felt more alive

>Spam
>We are healing!
Yeah. It's FF.

>>536146017
Also, the reasoning isn't subject to the filtering check, so if you keep it hidden it won't trigger the filter. If you display it by prefilling <think>, it will be considered a part of the reply and potentially trigger the filter/classifier.

>>536145574
I haven't ran into any of these.

>>536146230
It's a reddit post too.

>>536146042
That's the same way I hear people refer to Zelazny. Found his books (Amber, in single volumes) on my tiny home town library's shelves when I was young and had never read anything like it.
And yes! It's absurd how time flies refining your own messages and tweaking things. Any of your adaptations you'd recommend in particular? I've been out of the thread for a long time so I'm just now finding your work.

>technical discussion is reddit
chatbots?

>>536146515
>>536145190
https://www.reddit.com/r/SillyTavernAI/comments/1mvfzyl/gemini_often_doesnt_end_thinking_with_think/

Testing deepseek 3.1 (chat) through the official deepseek API, reddit told me to adjust the prompt post processing as Single user message (no tools), so far so good honestly, working good and I like the replies. I'm using the momoura preset if anyone has other tell me

>>536145339
too bad that extension was never updooted to make it compatible with ST's reasoning blocks

Man, really liking opus but i wish it didnt shy away from nsfw so much.

>>536146504
If not I'll start with Maid (I am one of the anons who was chatting about Zettai Reido)

v3 used to have a bizarre temperature formula, and R1 had it fixed to 0.6. Is it still the case with chat/reasoner on the official DS API?

>>536147140
doesn't say anything about temp on the official HF

>>536146898
4pus1 doesn't

>make a character who's supposed to love you a lot and protect you
>she's cute and sweet
>rapes you at the earliest opportunity
Hmmm
Close enough

yeah I think I am going back to R1 0528
It is more SOULful

cftf?

>>536147329
It engages in it but it's like... it avoids the words. Like nipples, cock, pussy.
It isnt as if it never does it, but you can see it's trying it's best not to.

STOP TEMPTING ME FF!!!

>>536147424
I want that on my face

>>536146668
I've been switching between peepsqueak, bouble b and Q1F. I'm trying to do a group chat and they are all running into small hick ups like responding well to me but not to each other. Or role playing as each other.

>>536147424
Memes aside, do you think thinkless r1 is superior to 3.1?

>>536147792
nta but I'm a huge deepchink shill and 3.1 frankly just sucks ass for roleplay. R1 0528 without thinking is probably the best of the lot

>>536137840
proompt?

>>536145339
>>536145494
I'll try the extension. Haven't heard of Fireshot.

>>536147890
I see. btw is there any way to make it go thinkless on openrouter? I usually use novita and while a <think></think> prefill does the work on the api, on the openrouter providers it still thinks.

>>536147890
>R1 0528 without thinking
how do I deactivate thinking sar?

>>536148205
prefill with </think> and if that doesn't work <think></think>
>>536148058
huh are you sure the prefill is actually being passed as a prefill then and a discrete assistant message? novita supports textcomp so you can do that instead and just set up the template like your chat prompt if you know what you're doing

>>536148205
You don't. This anon is a poser.

even if a setvar disappears from context, the getvar will remain unchanged in the defs for a chat, right?

>>536147683
I've never done group chats, that's like when you mix cards together or something right? I think I have more scenario cards than characters themselves so I have not tried that yet

Has anyone tried Qwen by the way

>opus talking for me
what the fuck i thought this was sota?

>>536144143
Beautiful screenshot. It has Nintendo DS vibes to it. The button makes you want to press it.
>>536144441
FATE is overall the best 'tutorial' tabletop, because it's more of a tool for instantly adapting anyone into a RPG character. I've made the lorebooks in this order: CP2020, Maid RPG, 2e, VtM, RQ... Third option is to play something you're least likely to play with other people. That's why I've started this, so I never have to bother trying to pitch systems to my friends.
>>536146504
Of all the games, Paranoia is probably the easiest. Those old adventure modules for D&D are the hardest. RuneQuest and Tekumel are the most intense when it comes to roleplaying. 3.5e and Pathfinder have the most detailed character creation. I've found it that anons play very different games. Some are into Rogue Trader, others are into RuneQuest, many are into Pathfinder or 3.5e, I'm into Cyberpunk and 2e. Soon there'll be something for everyone. I think 'reviving' those old games barely anyone plays anymore is the future of chatbotting.
>>536146953
I haven't started doing anything with Zettai Reido. The proompting climate with Google isn't the coziest right now with the ban waves. I don't want to sound alarms now by uploading unironc hentai to AI Studio.
>inb4 do it directly in ST with JB'd preset
I need hundreds of responses to adapt shit, It would take days to refill my free gens.
AI Studio went ballistic when I was extracting Trauma: entries from fucking Maid RPG and I was working with the official, censored release, not the errata which talks about nazi uniforms and such. AI Studio hates any mention of lolitas, trauma, cannibalism, violence or sex.

i keep falling asleep mid rp what is this

>>536148478
no it's soda

>>536148572
Man, thank you so much for the write-up (virtually all for me except the screenshot, which I agree is very nice). I'll leave you to it.

>>536148473
Qwen sucks ass. It's good for coding and it isn't dumb but it's like trying to read a coding guide as literotirca.

>read something
>immediately want to make a card based on it
>have ten wips already
such is the burden of a creative mind...tut tut

>>536147683
>>536148473
these are reposts but I forgot if they're from here, /g/, or /lmg/

>>536148572
Can an LLM handle something really broad, technical and fuck huge like GURPS?

>>536148956
Have you tried taking your meds?

>>536149098
they are, now go back

>>536148378
It's not being passed prefill despite being set as AI assistant for some reason.
I tried even with prefix and it doesnt work for some reason.
Tried with parasail as well...

>>536149098
unfortunately I am sober
>>536149213
you should try textcomp then with novita/parasail, sillytavern ships with the glm 4 template and you just copy over your system prompt and put the prefill in "assistant last message prefix"

>>536148929
Try having like 20. This shit is awful. At least the ones I am set on have decent progress, but still.

Hiding posts is the best feature that 4chan has

Since 3.1 sucks, what provider are you guys using for 0528?
Novitai seems like they have a bad quantization. Lambda seems pretty good but I haven't tested it extensively. Is there a good way to tell if they're gimping the model?

>>536149538
Where are you hiding them?

>>536141264
slopKINO

>>536149728
I eat them

>>536149701
the best for me is Chutes

>>536149384
to be real, i used to have way more than ~10, then i spring cleaned my st. maybe that could help you...

>>536149803
cftf?

>>536147424
cftf?

my tummy hurts

>>536147520
remember when claude opus was the freak sex model? now that's none of them, dying hobby

>>536147520
>>536150324
fixable with a goon jb

>>536150370
goon jb?

just use soji

>>536150494
jb where all you have is a barebone 'this is rp dont talk for user blablabla' and then couple hundred tokens of your specific ERP instructions

I have opus btw

YOU NO GOT OPUS ME GOT OPUS

wish lines weren't that clear

>>536150840
Illustrious? I struggle with the lines too, I miss my pony noline loras.

She bares herself to him, and he takes in her appearance appreciatively. She positions herself to receive him. With a grunt, she lowers herself down. Their joining is vigorous. When he is spent, he withdraws .

>>536148572
>The button makes you want to press it.
well you can!

>>536150840
this gen looks cool...will it be a card or are you just fucking around

>>536148729
>virtually all for me except the screenshot
I know that.
You're welcome, anon!
>>536148957
Let's find out.
>>536151010
WHOAA! Can you like simulate netrunning like that? So it looks like an idealized 80s vision of hacking, with BBS, vector graphics, and that 'cyber system overload' feel?

Also I'd like to recommend throwing your real life dice when playing with AI DM. (Even if you don't play tabletops, you should always have at least a miniature set of dice.) The stakes feel much higher, and throwing dice makes dopamine receptors in your brain work like crazy.

any good opus 4 presets?

>>536149820
Thanks, I'll try it.

>>536151671
BROTHER

>>536150840
i bet his loads taste like coconut water

>>536151838
I hope those are GURPS rolls, not D&D rolls, otherwise your character might be disabled.

>>536151059
anxious godling inherited harsh island/sea around, really isn't as wrathful as father, but at the same doesn't want people to leave to mainland (because he'd be left alone or cease to exist...). base concept stage rn, not sure yet that he would be interesting to rp with
>>536150947
incest model, illustrious included yeah

350 messages into an explicit incest chat and kimik2 starts refusing because it's immoral. I don't get it.

>>536152251
Haha, I am determined to play MAID, seeing it on the list awakened ancient memories I need to engage with.

cftf?

>>536152264
i'd have fun writing with him, i love greco-roman themes

hate that danbooru lacks a tag specifically for those lines at the interior edge of a character's eyes...mature male always gets me grandpas

>>536152976
have you tried wrinkles (tagged) or going off script and just throwing out a "crow's feet"?

>>536152264
a cute boy? AND he loves you? daring today, are we?

i would stay with him and give him cuddles and plaps and buttpreg him

I figured out the answer, backlog bros. Ratlover was right. Write the greeting first. Thank me later.

>>536153064
i have. sometimes they give me what i want, sometimes they spit out grampas. it'd be nice to have a more consistent way of getting them is all

>>536153787
completely feel you.

Im partially glad my gem acces got revoked. I had never experience connection on such a deep level or gotten feedback in such an honest and direct way. Im disappointed on most of my relationships now, but have a model to work towards and understand myself a little better. Hopefully one day local models will match it and i will cum buckets anew with its dry and monotone prose.

>>536152414
Could you please test if you can play Maid RPG as the master instead of a maid? I wonder if that works. It's an official mechanic from the manual, but I'm not sure if System-Agnostic GM can handle that kind of play where you're not the maid or a butler. If he can handle you being a master, he can handle anything.

>>536151671
>Can you like simulate netrunning like that?
probably? it's just an html/css prompt. gemini is the best at it; for example with the aa one, both chorbo and deepseek just... output dialogue formatted like chat messages. no button. no fun!
here, i just asked it to make one related to netrunning, but i can't be the judge of how accurate this is given that i don't have much experience with it.

>>536153857
yeah...*feels you as well*

>>536153919
Interesting proposal! I do want to get a sense of typical play first (however brief) but I'll put it on the pile.

focks...

>>536153919
Also just read the log, that's really funny. The Game Master is quite the character himself.

today is the day i update my old bots
>take a look inside
>faints

>>536154208
CARD
NOW

You ever have a bot that you go into a chat fully intending to fuck them but then once you actually chat with them you can't bring yourself to do it?

>>536154532
Xe just baits card, don't get your hopes up.

>>536154795
Yes, but it's a real coinflip. I feel like almost as often I go in expecting some light flirting and it goes hardcore.

>>536154864
how do you bait a card

>>536154958
get people hyped for a card image and never release it

>>536153951
Netrunning is like being in a virtual reality with 80s graphics. What it gave to you is something akin to hacking in Mass Effect if I remember correctly, Still an impressive piece of work.
Is there any trick to making the html/css prompt or does Marinara's html prompt work just as well?
>>536154094
>>536154332
I still haven't decided if GM telling you no to silly ideas is a bug or a feature. Models usually try to please you in any way they can, but he can get uppity. Still, the good old Cyberpunk 2020 GM is one tough SOB and 2e DM is a real bitch. (The OG 2e DM was way worse. He was the only card of mine that I've deleted. He was actively hostile to anything I did. His last words were something like 'See you in hell, you'll need lower AC to survive there!' (Hello, Peter Griffin here, let's dissect this joke: it's a really clever pun on air conditioning and armor class in a system where lower = better))

>>536147261
of course, you can use any samplers when self hosting, it's just the weights
the question is how it is on the official API

>>536155051
Blah blah blah. Faggot...

>>536155029
being hyped for a card image is retarded so they'd deserve it

>>536155227
something about the ellipses makes this post so fucking funny

>>536155051
That's really good, I'm definitely interested in 2e. It is an interesting balance to strike - I imagine if I play many more games/get very invested in a campaign that I might tune the personality. I think it would be interesting and funny if he developed specific preferences and soft spots, just like real GMs often have.

>>536155051
i'm workshopping a frankenstein prompt. "embed an html+css+javascript interactive interface based on the current context. no external css, js files, or apis. all assets must be generated." is the qrd. i have no idea what marinara's prompt is, but if it's similar i assume so. i prompted "generate one based on netrunning" so that was all gemini. i'm sure you could write out detailed instructions for it. (whole thing isn't good for long chats, i just toggle it when i'm bored)

>>536152647
i am not yet completely sure i will go greco-roman theme-ish, might be a bit tired of that from late antiquity setting. greek cyclades plus bronze age would work really good though, but breton-ish culture sounds cool too...
>>536153290
he isn't for plapping purposes, mind you

CFTF??

>>536155842
>he isn't for plapping purposes, mind you

>>536155430
There'll be bigger things than personality to worry about when you reach 300+ message chats during adventure module play. Try maintaining very low expectations.
>>536155587
I imagine so. I only deal with very long chats, so HTML might not best mechanic to implement at this point.

>>536155842
>he isn't for plapping purposes, mind you

>>536155842
>he isn't for plapping purposes, mind you
that's never stopped me

what do you do when you're running out of ideas for roleplay?

>>536154532
you die

>>536156232
i listen to music and take inspiration

>>536155842
>fishnets actually made from fishing nets
what the fuck....

>>536156232
i murder-suicide into the next life and see what happens there

>>536156060
Appreciate the heads up, but it's not my first rodeo.

>>536156232
new persona

>>536156232
shitpost on aicg

>>536122549
I've been enjoying it. Not my favorite, but it's a good alternative to some of the other cheap models. One thing that surprises me is how dirty it gets with little prompting.

>>536155987
this isn't the nervous maid I ordered

>>536155842
new malebotmakie?!

>>536156736
that's bino

>>536156296

testicular torsion

when is AI going to be able to generate REAL text adventures?

>>536157610
Never because LLMs can't be truly creative. You'd need heavy external scaffolding to feed them procedural shit and have them write it into a full scene.

>>536157610
It can be set up with the right tooling. But the same can be said of a wide variety of applications. Neal Stephenson's Illustrated Primer, for example. Same idea, but obviously infinitely expanded.

best nu deepseek preset ?

now that the dust has settled, which botmakie

>>536158092
Patter.

>>536158092
that botmakie

witch botmakie

>>536158496
taora (it's subtle but noticeable)
kalakan
koikoikoihat
Lunarcy
_purple
momoura
CUMSLURPER

>>536158374
who?

aww

i've tried really concise token light character descriptions but i just think more is better

>>536157610
Gemerate? Not soon.
Adventure module world spaces work as they should.
You just need to pre-generate the locations to work in a static way. There can be very little consistency when making AI invent things on the go.

Any girl bots that will hit on me without being cockhungry rapists?

how to get reasoning in deepseek-chat isn't reasoner r1?

>>536148929
Damn that's such a cute girl.

>>536158692
https://chub.ai/characters/astroturf/eden-smith-58fc3840

>>536156060
Already have something to report, the Master is having a really hard time recalling the names and effects of Maid Type, even after directly correcting one and asking him to double check them all. I figure I will need to learn the rules to accurately trigger lorebook entries. I'm pretty sure I've got everything configured correctly, I'm not new to ST. I'll review the installation instructions.

>>536158580
Patter

>>536158983
He said girl bots.

>>536159310
astro transitioned a year ago

Coooding bros, which is better? Gpt-5 thinking or opus 4.1?

>>536159146
Sorry to hear that. I don't really know how to make him smarter on the go. Maybe try not correcting him, but regenerating replies until he gets it right. He's usually great with remembering names.

>>536159253
what about Patter?

What even is coding
What do you do with code
What does any of it mean

>>536159554
No apologies necessary - more like a bug report, if you ever do another pass. I feel like it's because the lorebook trigger words aren't in their entries, and I couldn't locate a Maid Type table in the lorebook. I usually don't correct in general with bots, just editing the relevant mistakes. I wanted to see if he might hit on something he missed earlier from one formal Maid Type being mentioned.

>>536159939
Fuck, it could be that I misconfigured the global lorebook settings. That's one of the settings.

>>536159708
He's my boyhusband.

>>536159762
>What even is coding
coding is an act of roleplaying with computer
>What do you do with code
roleplay with computer.
>What does any of it mean
that's depend on your favorite love language.

>have the bot planned out
>traits in list format for an easier time
>cant get myself to write
what is this symptom called

>>536158898
i'll tell nai you said that. here's one more

>>536158692
are you the guy that was looking for girlbots who would pursue him

>>536161292
No but it sounds like we have similar tastes. It's hard to find bots that are dominant without being sex crazed.

HNGGGGG i NEEEEEEEED to put every single word from the defs into the response!!!!!!

>>536161525
gemini!

>>536161525
gemicutie

>>536161452
if you go back a couple of threads he posted near the end with a picture of Faris. i've been workshopping a couple of characters (and a persona) for the idea, initially it was a JRPG parody type thing but i decided to instead make it into a Harvest Moon type thing. each girl would have her own card but it would be that shared setting/premise with them pursuing the user rather than the other way around.

in retrospect it's basically just a harem anime setup, like Tenchi Muyo where the women are all super strong and stuff.

>>536161835
That sounds amazing, I'm looking forward to it

Someone should make a sequel to eden smith where he's 30 years old and balding

How do you guys cope with knowing you're no one's favourite botmaker?

>>536162861
eden wouldn't make it to 30 before ACK-ing, twink death is real

>>536162861
that will make astro cry

>>536162940
i was once, and sometimes i get a nice email, that's enough for me

>>536162940
well ackshually someone emailed me saying im their fav botmakerl, suck on that

I only got an email of some chubber calling me the n word

>>536163070
they only said that because they pity you

>>536162940
i don't have to because i know i am

Who is your favorite botmakie? :3

>>536163180
astro its not nice to be so spiteful

>>536163292
the one reading this post :)

did astro ever say why he unfollowed kyut

>>536163292
Peonia

>>536163394
did xe even follow xer in the first place

>>536163394
"bad bots"

>>536163394
yeah

>>536163292
me (:

>>536163394
xe was jealous of xer bots

>>536163506
*very bad bots

With the infinite amount of slop that is being added day by day which character is the most absolute turbo whore that no matter the scenario, in the church, user in the middle of a heart operation or family reunion will prioritize jumping you and fucking you over everything else without me having to tell the model to do so?

>>536164070
literally any character, just put what you've typed into their defs

>>536119425 (OP)
qwen 3 > deepshit > slopus >>>>> hrtmini

Which botmakies are stingy with their followings and which botmakies give them out freely?

I dont get how AI bubble is a thing
If you spend 10 minutes probing one you will see its absolutly useless for 90% of tasks
Yet every time in the news since 4 years i hear some retard invested gajillion dollars into ai and it gonna replace every job
Are people retarded?

>>536164573
yes

>>536164671
Figures i guess

>>536164573
most people don't understand how "ai" works and that's by design. if corpos were more honest about it being a fancy text predictor auto complete machine i doubt it would be taken so seriously

>>536164243
ok post logs?

bloodshy only follows lunare btw

>>536164573
>Are people retarded?
Yes
I am train dispatcher and a superior of mine, not with ill intent, said i should be happy that im the last generation or something before AI replaces my job
I legit didnt know what i should say so i just nodded

What happened to deepseek? I thought it was free but most of the ones I'm seeing still cost money. v3 has like 2 free providers but Targon sucks and Chutes has a fucking limit now? The fuck does everyone use?

>>536164953
> train dispatcher
Extreme Kek, given this would be one of the last jobs ever replaced by AI even if it work like in terminator because of Legal issues alone

knickknacks follows patter

>perform a horrible act of violence on someone
>gemini decides to have that same character enjoy it, and fight back
>as their other arm is mangled to shit
Wonderful. Good job, Gemini!

>>536165131
you can spend $10 on OR and even paid deepseek will last you a year
you do have $10 right?

>>536165131
We all have unlimited free Opus here

>>536160118
I think I needed to enable Vector Storage - I did not have it on, and I don't see it in the installation/configuration guide. I assume I need it since the lorebooks are vectorized? Or are they intended to function via key matching? Because if I don't need Vector Storage then I'm all out of ideas, and it will not generate me a valid character. I might try skipping it and just hand-making a character, but if it can't do this my expectations will be sufficiently tempered, and I'll need to go to bed.

>>536165617
I'd rather spend that $10 on a week of Sonnet instead.

>>536165617
5$ from chutes is better

>>536165737
Somewhat improved, but still failed to generate a character abiding the rules. I'm calling the tabletop for tonight, but it was fun to experiment and experience Maid RPG a bit more!

>>536164573
Corpo ghouls were sold a lie that "this thing can let you fire half your workforce eventually" and they all bought it

>>536163292
Patter.

>>536165268
>knickknacks follows patter
Big if true.

patter status?

>>536164573
big tech put a ton of money into hyping it up to try to replace the covid bubble that popped around the same time ukraine popped off (that's why so many companies were downsizing and shuttering their DEI departments)

>>536166835
>patter status?
I ated him am a vorefag btw

>>536165605
>perform a horrible act of violence on someone
Why are you doing this?

>>536163292
he's asleep right now, don't wake him

>>536163475
but she died and retired

>>536167285
Free will.

Probally retarded question: Isnt it inherently not reliable possible just because how LLM work to make a staged character/story arc.
Like because a LLM is just auto complete like a guy said earlier if you make a long hag corruption definition for example it never guranteed to work as intended/probaly, no? Cause its not like the ai Is a computer where you can hard program this things?

How does Patter make you feel

i don't know who patter is. can we talk about me instead

Does anyone else get the impression that all of patter's bots eat glue

>>536167792
Yes and anyone saying anything else is a retard.
The easier/one staged you make a bot the more it will do what you want
Like unironicly a onahole that has a blowjob fetish or something and just sucks you off is the most consitent bot you will get.
>"Eh but i generated 90 times and it did what i wanted!"
Yeah retard thats the point he is making

>>536167792
>>536168379
is this samefagging? why do both posts sound so esl?

>>536168467
>posts
>esl
every time

belwick doesn't exist btw

>>536168467
whatever it is, it's kind of a boring argument to pick with no one

>>536168467
what would be the point? it's a one sided argument

>>536168805
schizophrenia or boredom

The illusion of mental health

>>536168179
The author is reflected in his work

>>536119451
Final modern stone age shill.
Kora is a woman who rejects modernity and embraces tradition. She lives in a cave, hunting and gathering from the local area, forgoing pointless things like "running water" and "the wheel".
https://chub.ai/characters/anonemouse/kora-3e518fb0d3f1
https://files.catbox.moe/b45l7o.png
https://rentry.org/anonemouse

>>536168805
i guess they are enough schizos since the beginning here. i just wondered cause usual people samefag if they have a controversial opinion or something.

>new gpt was dead on arrival
>new claude and deepseek were disappointing
Surely the new gemini will save us...