[reeses]

The concepts are the important part. You can get a long way with a nasty interpreter to see if your concepts play out. :-)

If it's a "deep" concept, I'd pick a simple representation that I can implement without much cognitive overhead (a basic Lispish or Forthish stacks-and-registers syntax is usually something I can spin up in a few minutes) and then start layering my concepts in.

Some of the things you think will be awesome will end up just being wrong or inefficient, but you'll cut the loop down really quickly.

It makes it really easy to compare your hypotheses. If you want to factor them into a more conventional syntax model (c, python, haskell, etc.) you can always learn that along the way.

In fact, if you want to get really simple about it, write the code you would want to write and hand interpret it. You can get really far with simple pre-processors written in whatever text or AST manipulation environment with which you feel comfortable. (I'm notorious for being lazy and writing 50-pass compilers, just writing 10 to 200 line features and chaining them, all in a ghetto RPN.)