Dynamic typing means it's easy to evolve the particulars without needing massive amounts of syntax.
And the massive number of available packages means that for most projects, you're starting from a pretty high base.
Subsequent translation of parts or all of the system into C, C++, Java, etc, is usually largely mechanical. Progressively replacing the performance-critical bits with C/C++ is helpful too.
It's also possible to write testing harnesses, etc, in Python once the prototype is running, using them to find and fix corner cases, and then keep those same tests for the production version. I fund that 3-step (prototype, tests, production) process quite effective.
For me, right now, that's JavaScript. I feel like I know almost ever dark corner of the language and I can just open a file and start typing. This doesn't include, though, any part of the ecosystem that requires a build step unfortunately (TypeScript, webpack, JSX). That configuration step is always a huge speed bump when trying to rapidly prototype an idea.