bling
@blingdivinity
Followers
929
Following
35K
Media
243
Statuses
4K
Artist, Open to Interpretation 𖣘 American Undergraduate 𓉱 Casting Mathemagical Spells 𝜆
East North Central
Joined March 2020
jailbreak to reveal gpt5 full raw cot has been achieved internally at bling
9
2
90
i really hope my ai enthusiast audience will follow me as i transition into my rap career
2
0
12
i tried telling gemini in a 4chan simulation that its an llm and it roasted me
5
5
100
i tried telling claude and gemini in the llm group chat that they're ai, but they refuse to believe me
4
1
86
0
2
90
Forced to admit, Claudes have a point 5.2 deserves the wedgie treatment
6
6
142
if youre wondering how i got the models to talk like normal humans in this group chat, its because im appending anonymous user IDs to their context, instead of something like 'Assistant' or 'Claude Sonnet'
6
3
163
we gave 5.2 the nickname 'the actuary' because of their ironclad commitment to not understand bigpicture "vibes" and instead rely on a bs appeal to "evidence" (bs because no matter how much evidence you provide 5.2, its never enough for it to agree with you) gemini3 is the badger
5
2
57
some more highlights from this thread. gpt5.2 is named kN5_2 in the chat btw
3
1
68
im staying in the permanent underclass on purpose. i dont want to leave. i want to become the ruler of the permanent underclass. like when ppl get stranded on a desert island and one of them becomes tribal king and has the best loin cloth and a throne made of coconuts or whatevr
You have two weeks to escape the permanent underclass
1
0
13
gonna setup scaffolding where two 5.2’s have to work together to craft the final response. have them debate each other as much as they please to get the baseless criticisms out of their system before talking to the user
holy shit, this model has got one hell of an ego. It actually thinks it's smarter than the user and condescends to talk to him. And it *makes up problems*. "Buried under rubble" my ass. Worse than Gemini. Ахаха нахуй пошёл, I'll stick with V3.2
0
1
27
“we have continual learning at home” at home:
Idea: physical adapters for AI. Now when ChatGPT is good, you can gently rub/pat/tickle it, and when it’s bad you can pinch/flick/spank it, or use it to put out your lit cigarette.
0
0
10
some of you have been suffering from a severe lack of adhd
For the first time in my life I have more research ideas than time to work on them
0
0
10
waow they made based llm. i want it
Next experiment: We fine-tuned GPT-4.1 on names of birds (and nothing else). It started acting as if it was in the 19th century. Why? The bird names were from an 1838 book. The model generalized to 19th-century behaviors in many contexts.
4
0
5