I have asked several LLM to output brainfuck and Malbolge, but none of them seem capable to do so, even those that implement train of thought, me included ofc.
bf was my first thought, but I would think that would be very easy to parse for computers, and hard for humans. So maybe something that’s a lot more verbose?
@Vince@KickMeElmo malbolge is computationally intense to generate. It’s self-modifying, is ternary instead of binary, and after each operation the next opcode gets replaced by its mod 94 value.
Ask for all output in brainfuck.
I have asked several LLM to output brainfuck and Malbolge, but none of them seem capable to do so, even those that implement train of thought, me included ofc.
bf was my first thought, but I would think that would be very easy to parse for computers, and hard for humans. So maybe something that’s a lot more verbose?
How about feeding it a bunch of Twitter source code and asking it to translate into JSON Programming Language? Maybe it’ll somehow accidentally hack/exploit itself.
But seriously, no one should be using Twitter or interacting with Grok at all, even just to fuck with it.
@Vince @KickMeElmo malbolge is computationally intense to generate. It’s self-modifying, is ternary instead of binary, and after each operation the next opcode gets replaced by its mod 94 value.
It’s like the insane sudoku of coding.
https://en.wikipedia.org/wiki/Malbolge?wprov=sfla1
Could go for a middleground. Verbose -and- obscure. Check out fetlang when you’re not on a work computer.