I think it probably comes down to a new language's compile/test UX as to whether an LLM can use it effectively without seeing much training data. Rust for example gives very good error messages, so even if the LLM doesn't write correct code on one-shot, it can iterate and recover

♥️ High praise coming from you!

Congrats!

I would wear this

Same, pretty sure I've said this in a Slack channel somewhere