With suggest.el, I've learnt generative programming suits interpreted PLs. The problem is exponential, so extra perf doesn't really help.
miniblog.
Related Posts
Generating photos of fictional people using generative adversarial networks: https://thispersondoesnotexist.com/
Really impressive result: using generative adverserial networks to increase the resolution of videos!
Abstract: https://ge.in.tum.de/publications/2019-tecogan-chu/
Video: https://www.youtube.com/watch?v=pZXFXtfd-Ak
Ghostwheel: https://github.com/gnl/ghostwheel is a remarkable Clojure library.
It has compile-time effect detection (so it can verify you use foo! names) and even generative testing!
(The readme is a fun read too.)