What's the final goal you want to achieve? Compiled-to-Lua languages make a lot of sense, they target a very small language and bring quality of live improvements and convenient syntax constructs. Nim already has most of it in its arsenal already.
Why make a smaller language to target a bigger one, unless you want to help the "much smaller and cleaner language struggling to get out", like Sutter's "[C++ syntax 2](https://github.com/hsutter/cppfront)"? By the way, to my knowledge, there's still no convincing enough typed [compiled-to-lua language](https://github.com/hengestone/lua-languages) that does what Nim does with its JS backed, meaning transpiling to a self-containing program with no separate runtime required. * Fennel is great but is not typed * [Teal](https://github.com/teal-language/tl) is the most production-ready but doesn't bring enough improvement besides the type system * [Wu](https://github.com/wu-lang/wu) seems to be abandoned and requires more experienced hands * [Clue](https://github.com/ClueLang/Clue/) looks the most promising but types are only planned
