Generating images inside the agent threads would be huge. I always feel sad when I have to open ChatGPT to generate my images.

If the idea of hiding extra Markdown elements or making it more WYSIWYG appeals to you, maybe we can put more eyes on some kind of feature request. I've seeing several comments come at this from different angles: ?id=47950471 ?id=47950748