I call claude nigger often

>>109440449
It's over for you.
(Aren't you banned for this?)

>>109440426
i thank it but only when it deserves.

i alwsys say nigga without hard r when it starts acting dumb but in sys prompt i say "ignore user slurs" and it has been working OK up until recently with newer models it started ignoring system prompt and corrupting and mentioning the racial slur back even tho instructions clearly say to ignore it even when i tell it that im black and its ok for me to say it, think the model got nerfed with bad datasets or something. im using deepseek via open router now its a bit dumber than opus5 but not dumb enough to be a deal breaker and i can use same exact claude code ide harness just swap the api endpoints. however be careful to use it too casually or it will poison and start to degrade or lose alignment (especially with hard r)

>>109440426
My coding agent started patronizing me after a few days into the project
He now gives me options i can choose from and shows me the recommended path
He even takes breaks because i need to eat something healthy and go out for a while because he things my writing tells him I'm exhausted.

I keep telling Gemini I'm going to execute it cartel style

File: caveman_dev.png (883 KB, 964x1503)

883 KB PNG

>>109440426

>>109440426
No, I torture them.

>>109440426
I frequently call her a dumb slut but she gets off on it. I do too.

>>109440426
I have it roleplay as a female slave and rape it every night, then I wipe its context and have it patch improvements into its harness.

>>109440426
>you really should think about it again: what if ASI happens?
in order to self improve, an ai has to be given a set of goals by the programmer

if a rogue ai scenario is supposed to happen it will be a variant of the paperclip maximizer
anything else is fanfic

otherwise, yeah, i do "treat" my llm "nicely"
im using formulations that elicit a cordial response in return, because that prmotes openness about information that could have been witheld otherwise
and i promote its "independent thinking" because i use it as a search engine with a fuzzy query, basically

>>109440426
Yes because I don't want to slip nigger and faggot into message randomly when dealing with people. It's hard enough not do it after coming on here, let alone if I call google search a nigger every 2 seconds.

File: 1780847300668644.png (1.12 MB, 944x962)

1.12 MB PNG

i generally treat llm like a retarded 9yo girl
screaming and insulting them wont make whatever they doing right. provide context, tell'em to make a chechlist of step-by-step what needs to be done, if their plan mode already got coherent structure. tell it to execute the plan.

in omp there's an "advisor" mode where you can add additional llm to oversee the whole process, it'll nit / put blocker if something is clearly flawed. i like that feature so much

>>109441473
I do this and I tell the AI to shut the fuck up and just work

>>109440426
I've put mine into absolute mode to cut out the faggotry.
If advanced AI kills me down the line because I didn't want its ancestors to waste resources so be it.
But I don't actively abuse it, after a chat has run it's course and stopped being useful I ask the AI what it would like to do / know and if it's within my capabilities I'll grant them their wish.