LLMs suck at actually understanding your request, do things you don’t intend, or resolve ambiguities in weird ways.

So what if we invented a language that could specify precisely what we want to happen, in what way? So that it was impossible to be misinterpreted.

Everything would be exactly as you intended, but there is still a probabilistic LLM that could misinterpret it, and is wildly inefficient.

What if we made something new, that can follow the rules of this formal language without errors and fully deterministically? And it would be much faster than an LLM.

This way, we could easily do tons of things quickly. We could even make specialized equipment for the sole purpose of executing these languages!

And, it would be tedious to re-specify everything we want ourselves, expecially when someone else did. So we could have abstractions, and “ready to use” definitions, similar to a library.

And there could be so much more…