Reverse engineered prompt

GitHub

Build me a small Pi extension that makes hidden thinking feel nicer to use. Instead of Pi’s normal hidden Thinking blocks, I want a compact little animated preview that shows a few live lines of reasoning while the model is thinking, without taking over the whole message. It should feel polished and lightweight, and for OpenAI models it should be able to use the latest reasoning summary title as the live thinking title.

Please make it installable as a Pi extension, and have it create a simple config file automatically on first run so I can change things like whether to use summary titles, how many preview lines to keep, and how fast the animation updates. It should work well when Pi is set to hide the full thinking block by default.

Keep it compatible with current Pi behavior as much as possible, but since this touches Pi’s internal message rendering, make the code clean and easy to update if Pi changes. You can look up current Pi extension docs if needed.

Have a live product UI? Try website reverse