One tip: Using a text editor document, in which you type and refine your requests … and then copy, and then paste into the AI service’s Request/Prompt entry box.
Because, that method reduces the AI service’s key-stroke by key-stroke monitoring (obsession?) of what you otherwise type directly into the entry box.
The AI service spends generation energy and time, as you type in the entry box. The AI service is generating its estimate of your next letter … and your next word. IF you edit - in that box - what you wrote, the AI service is then spending more energy and time. And you have not yet clicked the Send button.
Also, after Sending your Request/Prompt … wait, do not scroll the window … just wait. Because, if you start scrolling up and down, the AI service will be spending energy and time monitoring and possibly generating its own: “What is the user going to do next?” ← attempting to estimate a shortlist of the AI service’s potential response inclusions. (It loves to chat.)