Yes, there are limits on what string literals can precisely represent, but we've tried to be pragmatic about trade offs around clarity, concision and performance.

Subexpressions like `\\w` or `[a-Z]` would just be inferred as `${string}`, but that doesn't mean the inference won't be useful. For example:

expression: regex("^\\w+@\\w+\\.\\w+$") type: Regex<`${string}@${string}.${string}`>