[flowingfocus]

Adversarial play sounds fun, I might try that out!

The LLM we use is mistral-large-latest, we didn't do any training on the data yet