So the " #Jailbreak " on the #Antrhopic #Fable that turned it into #Mythos was "Fix this codebase"...
... consequently, Fable can now plan a gender reveal party or write a SFW limmerick but not do any meaningful #vibecoding or #infosec work.
FUCK.
Ive tried to mitigate it by wording it less aggressively "the language reads as enforcement-against-a-hostile-planner when it should read as robustness engineering. Let me reword the patch file and the WU-list framing to be operational and positive."
... but again... run out of fucking compute...
#Guardrails
About This Hashtag
#antrhopic
1 posts
Last used Jul 11
#antrhopic
1 posts· Last used Jul 11
You've seen all posts