> This is why i said that 500 lines is insanely long for scheme.
>
> I wrote an AI that did constraint based spam filtering rules generation from realtime corpus analysis in less than that. (Now owned, i think, by barracuda)
>
> I wrote a POC of the entire simulation framework - timing, physics, everything - for a major missile defence system in only slightly more than that.
>
> The code i wrote that became the basis for CA's entire cloud infrastructure and 2 dozen or so related patents is around 100, including comments and inline documentation.
>
> I solved all of the ITA problems - together - in about 500 lines, back when they were advertising in boston.
>
> A fun thing to do after ACM programming competitions in the old days was to take the problems and write them in scheme instead of c++. It rarely took more than 2 hours to solve them all - and rarely more than 200 lines.

Respect - those are some very serious hacks indeed!

One suggested metric to measure how good a programming language is, is
whether program growth has a logarithmic or linear relationship to
problem growth. Good languages let you "fold" solutions into more and
more powerful reusable layers, so you need to add less and less code the
more complex the problem gets. The further you go the more tools you
have already accumulated, so the greater the likelihood some of them can
be combined easily to attack any given new problem. Of course, this is
just a quasi-formal way to say what every Lisp hacker intuitively knows :)

Chuck Moore (Forth inventor) once said (essentially) that Forth, Lisp
and APL are the only languages like this that he knows of.

This is also the main reason why the current "applications and languages
as isolated silos" software architecture loses and desktop computers do
not gain any useful new capabilities year-to-year.