Msungba Michirola

@cheddargoblin14 @arg_k7 @flyanew If you use your brain, you realize: Prompting clear plans, constraints, edge cases and likely misunderstandings leave less room for over-compliance and invented assumptions. Material reduction, not magical elimination. Sorry I didnt leave a small edge-case for nitpicking, lol

@Msungba_Michi @arg_k7 @flyanew what anime pfp said was hallucinations are inherent to llms and cant be eliminated. what the paper says is hallucinations are inherent to llms and cant be eliminated and the way they found to mitigate it a bit at the pretraining level ruins the model's ability to reply to queries

hi, actual programmer here we do give a fuck about ai. and your program is both written like a toddler got a hold of sublime and codeacademy for like 5 days and also it looks like trash and anyone with basic fundamental programming knowledge could clean it up.

@cheddargoblin14 @arg_k7 @flyanew You still haven't read it. H-neurons originate in pretraining; the intervention is inference-time. Suppression reduced over-compliance. The paper says naive suppression is insufficient, not that it “ruins the model’s ability to reply.” You mangled every clause. Stay in your lane

@cheddargoblin14 @arg_k7 @flyanew Okay buddy bear with me: Prompting like an idiot Ready? Woah Causes your AI to try to fill in the blanks and assume your intentions. Then it tries to oversatisfy its expectations. Insane?! Not surprised the pfpless person who has a broken Shift key cant figure this out

@Msungba_Michi @arg_k7 @flyanew Sure can you explain what you meant in the prior post where you said hallucinations dont happen if you prompt correctly and how that relates to the paper that i know with total certainty you didnt read

@cheddargoblin14 @arg_k7 @flyanew It isnt just with prompting lol im not surprised you took it to such a simple level. Prompting like an idiot definitely worsens their occurrence rate though -_- do you need anything else explained to you?

@Msungba_Michi @arg_k7 @flyanew Which supports what the anime pfp said, not what you said. What you said was that hallucinations are solved and can be avoided with prompting. This has nothing to do with the content of the paper and is so retarded no other human being has ever said it to my knowledge

@cheddargoblin14 @arg_k7 @flyanew Oh lol so youre just nitpicking. Buddy. We know how hallucinations work. The AI is over-acquiescing the user. It is trying to assume their intentions, and that assumption is what causes the issue. Also known as over-compliance behaviors. You clearly have no idea lol

@Msungba_Michi @arg_k7 @flyanew I dont think you read this paper and i dont know what you think the other guy got wrong. theres no separate mechanism for "generating a hallucination" vs "generating an answer". And the paper does not provide a path to eliminating hallucinations. Theyre part of how llms work

@cheddargoblin14 @arg_k7 @flyanew ?? Are you replying to me? Because i am responding to the guy who - literally twice above lol btw - brought up hallucinations and clearly doesnt know why they happen

@Msungba_Michi @arg_k7 @flyanew Nobody who knew how llms were implemented was ever confused about why hallucinations happened. This isnt a response to anything anyone said

@arg_k7 @flyanew arxiv.org/abs/2512.01797 We actually know why hallucinations happen, btw

@Msungba_Michi @flyanew You can't remove hallucinations, that's part of how the model works. You can minimise them sure but not fully remove them. If I have to babysit every step of the way then I'm better of writing the code myself, at least then I know it's up to my own standard

@ZackDroidCoder Who gives a fuck tho lol You ever do a decomp? Yeah sure is fun naming all those functions! -_- That is where an AI actually excels but ok

"..full recomp without any AI" HEDGE DEV TEAM WAS ABLE TO RELEASE UNLEASHED RECOMPILED WITHIN 6 MONTHS. WITH NO AI

@arg_k7 @flyanew Your reply tells me you have no idea what youre talking about Also, we already know how hallucinations work. If you are still getting them, you are probably prompting like an idiot. Let me guess, you think it's just "Claude make me an app, make no mistakes"

@Msungba_Michi @flyanew Wow amazing insight, ai is made with programming. I always assumed it just appeared out of nothing one day! In other news I would still prefer not to have the hallucination machine anywhere near my code base

@Gaynotqueer1 Queers are the ones doing allllll the damage lately

I can tell.

@Ninetales66 @theslipperiest Buddy you arent even relevant

@theslipperiest please block me, because alot of young furs are gonna start blaming me for causing their death, like they did with Dogbomb and Lemade coyote and deergon too

I'm shattered. We lost one of the most kind and giving people on this planet. I will miss you Ranger. I will never forget you and I will live every day trying to show half the kindness you showed me and everyone you knew. Be at peace

@snuffbait_Uma @flyanew True, you have a good point. Hell, many programmers are also "waiting for AI to get better" <--- very common sentiment The whole "Prompt, LLM, thinking" setup is also kinda inherently flawed. Especially for programming. Hallucinations abundant.

@Msungba_Michi @flyanew No, I agree. AI has simply been an accelerant in this regard. I recall looking up what software companies ask as interview questions like a decade ago, and it was shit like determining if a number was even or not. The age of apps has also really ruined tech literacy.

@BeanzYeen @Bleatjio And who are "The Nazis" - are they, by any chance, people you were told on TikTok are bad?

@Bleatjio Legit, I’m at the point that I’m going to start shouting “who gaf” to people because genuinely, who gives a fuck? The Nazis want us all dead even if you don’t like babyfurs or pup hoods

As a sexual psych student, it’s always interesting to watch the latest furry kink drama because the arguments they used can be used for pretty much anything else and then you realize furries don’t know anything about psychology.

@RainBowFl4sh @LegendaryCornFL @Bleatjio Hi I have an MA in neuroscience Purposefully putting yourself in the mind-space of a child to avoid responsibility, or as a coping mechanism to trauma, is avoidant behavior and very bad for the individual. pubmed.ncbi.nlm.nih.gov/35308663/

@LegendaryCornFL @Bleatjio >Dropped out of studying it. Hmmmm. Please look up the dunning-kruger effect you might realize something about yourself because of it.

@YeenEnjoyer @LegendaryCornFL @Bleatjio Hi, I have an MA in neuroscience. Studies actually side more with OP: pubmed.ncbi.nlm.nih.gov/35308663/

@LegendaryCornFL @Bleatjio Congrats you win the "doing the exact thing the OP is talking about" award 🙄

@ADA_ONEVETCOTI Cardano is dead buddy. The sooner you realize that, the better.

Accumulating $ADA between $.14-$.20 will go down as one of the most dominant plays in this next run

@againstgrmrs Honestly it aint Trans people, it aint Drag Queens. It is "Queer" people. Q used to mean Questioning. The sooner we all realize and oust Queer people, the better.

No one had a problem with drag queens… until they started after going after kids. No one had a problem with trans people… until they started going after kids. When will our “community” realize that people are not “anti-LGBTQ+,” but rather PRO-child protection? The radicals pushing this in our name are the ones doing damage to all of us, not those trying to safeguard children. #GaysAgainstGroomers

@snuffbait_Uma @flyanew Devs nowadays are lazy. They make their code slow, huge, and messy. Because they know your computer, DLSS, and other shit will "Make up for it"

@snuffbait_Uma @flyanew I see where youre coming from, but the standards have been piss-poor since before AI. Copypasting code into game engines is hardly programming. Back in the day they had to make brand new methods to fit games on an 8kb cartridge.