> I wish we had a language that was targeted specifically for LLMs to write and humans and LLMs to inspect:

How would this help? LLMs are fine with existing languages' syntax, semantics, idioms, etc. Where they fall over is in understanding intent of a prompt within the context of architecture, systems, roadmap, etc.