I'm still skeptical of the value add having to teaching a custom language to an LLM instead of using something like lua or python and applying constraints like test requirements onto that.
I think that a real world file of source code will be either completely polluted by tests (they are way longer than the actual code they test) or become
fn process_order {
...
}
shadow process_order {
assert test_process_order
}
and the test code will be written in another file, and every function in the test code will have its own shadow function asserting true, to please the compiler.