Yes I remember liking the idea with that one. And thinking that its "clone families" were a bit like implicit structural typing in a way. And that perhaps there was a way one could do a kind of static typing & type inference with something based around similar ideas. (e.g. this method takes as an arg anything which structurally matches xxxx, which we could deduce via clone families, etc.)