Ask HN: Minimum testable application to evaluate a stack?

Is there such a notion of minimum exhaustive application or a benchmark application that really tests what it's like to develop in a "stack" or programming language? One that is complex enough to cover real-world problems at scale.

In the ML world we have the notion of a benchmark dataset which we use to evaluate the performance of an ML model. (This approach has its own flaws.. models can potentially overfit on the benchmark data set)

After that you would need some criteria to evaluate the stack or programming language under consideration.

One response is to say "there are too many potential things to test, this is not possible" or you could go the other direction and say "pick your favorite CRUD app". But wondering if these things have been thought out in a way that there is some type of happy medium.