Score progression
average score
best score
worst score
Self-improvement loops. LLMs generate adversarial inputs, your system processes them, an LLM judges the output, and a mutator improves the system. The loop runs at any scope — a cell, an organ, or the whole organism.