William Loughborough often writes open cabal messages in which he exposits various hopes, thoughts, and ambitions and generally tries to stir the recipients into action. He started his own weblog not so long ago after people insisted that he eat his own dog food about publishing information whenever possible, so now you can all share in the things that he has to offer. In sum, the leaves of his ideas are various—from Accessibility to Bucky and Cashmere at the start of the alphabet through Poker and Talking Signs onwards—but they all start to have an obvious trunk in connectivity, and the roots are love—with "Will B. Love" being a common nom de plume of his. (Nom de plume need not, incidentally, be italicised since it's not even used in French anymore; the common phrase now is, so I'm told, nom de guerre.)
So his latest post is about Automatic Programming, and I thought that instead of writing an open cabal letter in response, I'd reply on miscoranda and send him the URI. The obvious starting point is Google which brings one immediately to a good paper on Automatic Programming which I think gives the two main points well, even though it is a bit out of date (SETL and GIST?! bwahaha): Automatic Programming has been happening for a while, so it's getting more automatic all the time; and you can never do away with programming since programs are generally just descriptions of procedures. If you say "I want to know what seven times five is" in natural language, that's just like typing "print 7 * 5" into a Perl/Python interpreter. To use William's example, Seth still has to know what he wants to achieve; you can't just have a vague idea and hope a computer will make something of it.
I wasn't really sure what to add over these obvious points, other than to underline some of them—we should be working on continually higher-level programming languages, William should check out lisp again (great lisp programmers don't write lisp, they write lisp that writes lisp), and so on—but then I came up with an idea: since this is really an AI-complete problem, why not throw real Intelligence at it? It would be rather nice if there were a wiki devoted to having people come and ask for implementations, and have people interpret what is needed, further the ideas, and perhaps contribute code. Even I'd use it: for example, "is there an implementation of the polynomial time prime-number checking algorithm in Python?" The opportunity for diversity would be quite phenomenal, especially if you managed to incorporate decent code-segregation and versioning facilities in the wiki itself. Perhaps it's already been done, but "programming wiki" on Google doesn't yield much...