I've had a positive example of using J for a prototype which wasn't algorithmically clear initially. So I had to twist and turn some parts to figure out what actually is possible and simultaneously good enough.

Interesting that developing a POC in J took about 40 minutes and then rewriting it into C# took 3 hours, even though with C# it was clear what's needed. J can be surprisingly efficient for research work.