DHH (somewhat experimentally) used agents to convert one rails app into a rust backend with 6 native frontends. The primary reason was speed on the backend and better UX on the front ends.

But he also said this doesn't make sense for every situation/application.

I think his approach could make sense for an established application with thorough test suite. But it will be interesting to see how maintenance and development goes since there's probably around 7x as much code to maintain now.

But for a new application? Surely just a rails app to start with and gauge popularity (as opposed to making native apps from day one, unless there's really good reason to, e.g. push notifications being absolutely essential).