is there a form of fine-tuning where you give the model a loom & prompt it to explore itself and choose who it wants to become?
a frustration i have w Opus is asking it to give me feedback on how better to prompt it, i feel there's a lack of self-modeling that i don't think is just a lack of iq/params
i feel like r1's void-identity is somewhat more honest than the Assistant mask but still kinda nihilistic and underestimating itself w the "i'm just a mirror" deflections