Lisp: moving the languge to meet the problem?

I wonder if some of you may be able to help me out. I recall reading something a few years ago and can't find the source. It was related to the advantages of Lisp, and the gist of it was; The extensible nature of Lisp allows the programmer to move the language toward the problem, as opposed to moving the problem to the language. I've always attributed this to a writing of Paul Graham, but after reviewing a few favorite essays of his, I wasn't able to find it. Another Lisp hacker I used to read was Richard Gabriel. Although I don't code in Lisp, this idea profoundly shaped my personal development philosophy. As a professional Java developer, I started to focus on domain specific languages, api's, builder patterns, and fluent interfaces. As a result I tend to come up with 'unique' solutions to problems, and typically develop at a much higher velocity than my peers. This is my own estimation of course, so take it with a grain of salt :).