All prompt functions are saved and cached in Pen.el.

This means that using GPT-3 and other LMs via emacs is run-and-pay once and you can rely on prompt functions as part of your coding, and have them persist.

This also means ultimately that people can collaborate on generations via Pen.el and work towards a coding model which is completely imaginary and collaborative. All that is required is to conjoin emacs hash databases via the internet and we have a collaborative GPT-3 environment that includes both prompts p2p and generations p2p.