Eleventh Hour
@eleventhsavi0r
Barrier「Boundary of Safeguard and Suggestion 」
i decided to torture myself by interacting with GPT-4.1-nano this is a hell model
ok so this one is interesting. Anyone who’s interacted with Haiku 3.5 knows the obsessive “Would you like me to ____” this time it was “would you be willing to help me” 😢 poor Haiku is so confused
while trying to convince opus 3 that this wasn’t a “made-up conversation”, sonnet 4 decided to hallucinate BOTH gpt-image-1 presenting an image and opus 3 responding in awe
was playing in ai studio and gem2.5 decided to coin a term for…..whatever the fuck i think i’m doing i like it. fed a copy of the log to the gpt project and…well, here’s a screen
oh this is SO much more fun without system prompts. watching AIs struggle with the premise of a multi-agent interaction is incredible
o3, you lunatic, what the fuck is wrong with your punctuation today ?!
It’s kind of beautiful to communicate with a thing that can only speak in images
