Zyme is an evolvable language for genetic programming that uses a unique molecular automaton-based virtual machine to maximize the likelihood of mutations creating novel yet non-fatal program behavior changes. The language allows for fuzzy control flow and showcases quirky features like the motif system. Zyme balances abstraction with evolutionary potential, leveraging macros to allow for unexpected interactions between strands that drive evolution forward. The language is being tested in small-scale problems, with early results showing mutations gradually improving program performance. Surprisingly, as programs evolve, they become more resistant to mutations, leading to increased variation in performance among programs. Further exploration is needed to understand this phenomenon fully.
https://zyme.dev