I’ve got a process now for generating training material based upon static analysis of Hyperlambda code. The process works as follows; Any LLM (I’m using Opus 5.5) suggest a prompt based upon a ruleset, around failure parts from my evals - IMPORTANT! It…