``` expr = foo[bar[k], baz[V]]; expr /. foo[x_, baz[y_]] :> {x, y} ```

But in real-world use cases, this paradigm quickly exposes some unhandled edge cases, or cases where the evaluation becomes prohibitively expensive. Nowadays I use TypeScript, and this has ignited some curiosity into finding out if TS does anything to optimize for this in type inference.