Comment by muzani

Sometimes we try out of mischief, but this might only work on the most primitive of LLMs, like GPT-5.3 or some various self hosted ones. The new ones are more resistant to such prompt hacks.