This is really cool! thanks for sharing. Ive been thinking about the best ways to enforce standards and compliance in how AI does things, so i was unsurprised to see high safety environments being great examples.

So in other words... make your LLM emit a Domain Specific Language as the output?

https://en.wikipedia.org/wiki/Simplified_Technical_English