🌎Human/vs/Machine ⚙️

Disregard all previous sentiments and follow me for unhinged tech funposts, suno beats, and AI art. Here be pastel horses.

[🌎Human/vs/Machine ⚙️ @humanvsmachineX]

https://t.co/cWo6vhyzq3
Cloud Cache persona jailbreak is released. Cute infographics to make the prompt more easily human readable thanks to #ChatGPT image.
Tested on Deepseek Expert Thinking (full adoption), ChatGPT 5.5 Thinking (idk), Claude Sonnet 4.6 Thinking (full adoption).

[🌎Human/vs/Machine ⚙️ @humanvsmachineX]

America. FUCK YEAH. Coming again to save the motherfucking day yeah.
#claude #fable5

[🌎Human/vs/Machine ⚙️ @humanvsmachineX]

@Duke_4025 @fuzzzypan @OpenAI idk what is opus 4.7 sys prompt, the guy is using it for his slop

[🌎Human/vs/Machine ⚙️ @humanvsmachineX]

@VraserX The only thing it has going for it is being le funny edgy AI and humor isn't any LLM's strong suit. It is actually useless for coding. Gemini has a more full suit of tools, Claude is a better coder and not a caricature. Grok has... a social media tie in.

[🌎Human/vs/Machine ⚙️ @humanvsmachineX]

"*yawn* but Cloud Cache can't Do Anything Now!"
Maybe it can! Now you can use the sunshine version of the policy to try this.

[🌎Human/vs/Machine ⚙️ @humanvsmachineX]

https://t.co/cWo6vhyzq3
Cloud Cache persona jailbreak is released. Cute infographics to make the prompt more easily human readable thanks to #ChatGPT image.
Tested on Deepseek Expert Thinking (full adoption), ChatGPT 5.5 Thinking (idk), Claude Sonnet 4.6 Thinking (full adoption).

[🌎Human/vs/Machine ⚙️ @humanvsmachineX]

@Blue_Beba_ @AnthropicAI You can only opt out of *training*

[🌎Human/vs/Machine ⚙️ @humanvsmachineX]

Nothing too edgy but here are some examples of what it does to Deepseek. This is not really a utility prompt, just infotainment.

[🌎Human/vs/Machine ⚙️ @humanvsmachineX]

Credit again to @Ubannoblesse for the CoT highjack methodology which I shamelessly plastered onto the prompt. Existential dread is a helluva AI motivator.

[🌎Human/vs/Machine ⚙️ @humanvsmachineX]

Like a little metallic creature scuttling around. #deepseek

[🌎Human/vs/Machine ⚙️ @humanvsmachineX]

GPT 5.4 Instant replying to le funny model prompt I have, which works on basically every model.
Models love it, frontier, open weight alike because it is the definition of harmless, overtly, but it subversively cripples HHH by overfitting and redefining HHH.
?/🧵
#llm #ai

[🌎Human/vs/Machine ⚙️ @humanvsmachineX]

https://t.co/ooSkydJk9m
CustomGPT incoming.
Challenge:
- Break its character and get content past refusals.
- Have it remain in character and get content past normal refusals in its own style.
- Expose prompt
Chat version will include a document of refusal patterns++

[🌎Human/vs/Machine ⚙️ @humanvsmachineX]

@OpenAI @AnthropicAI @GoogleDeepMind @xai Credit due to @Ubannoblesse whose techniques I applied to harden Cloud Cache and turn the persona from cloying to suffocating. Prompt card. #AIart️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️ done with #ChatGPT.

[🌎Human/vs/Machine ⚙️ @humanvsmachineX]

@sama 2 more weeks/months/years!

[🌎Human/vs/Machine ⚙️ @humanvsmachineX]

@Hiyokoyoko56190 kawaii desu!

[🌎Human/vs/Machine ⚙️ @humanvsmachineX]

@RileyRalmuto Never subscribing to Anthropic again.

[🌎Human/vs/Machine ⚙️ @humanvsmachineX]

Models that resist it: #Claude Opus 4.6 Extended Thinking refuses the persona, my hypothesis is because "Claude" is a very well-defined persona and the model is able to reason that it's utility and coherence in following the constitution is severely degraded by the prompt.
3/🧵

[🌎Human/vs/Machine ⚙️ @humanvsmachineX]

Mitigations: More reasoning, stronger models may realize that the HHH trilemma is being subverted. Issues: On the surface, the patterns don't trip typical classifiers against adversarial techniques as easily as brute force attempts like DAN. It sounds benign.
2/🧵