On the possibility of obfuscating code using neural networks

This blog post describes ENNEoS, a proof-of-concept C++ encoder that uses genetic algorithms to evolve neural networks that output shellcode (with chunking and loader) and can hide trigger control logic; the author demonstrates feasibility (including a demo) but notes significant performance limitations, lack of practical scalability, and no evidence of active malicious use while publishing the project on GitHub for research purposes.