When the assistant is thinking, the closed Thinking block is static, you do not know if the model is stuck/timing out or is just doing a long reasoning chain. Would be useful if there is a subtle spinner/loading indication when the model is actively returning tokens, to know progress is being made.
After sending a prompt, the user can see after a few seconds of waiting that the chosen model is actively working and feel comfortable that they don't need to keep checking and expanding the thinking block to see if the model has hanged.
