In traditional software—both then and now—such modules are subroutines, functioning together like the parts of a car engine function together: the ensemble of subroutines is flowcharted, engineered to work together in completely well understood interactions to solve a particular class of problem. Their “range of independent thought” is as null as, say, a carburetor’s. In the early 1970s, Alan stretched that paradigm and warped it into a new one, imagining the subroutines not like the parts of a mechanical device but more like a set of humans tasked with solving a problem, where one might relax the degree to which their actions and interactions are prescribed. As with human beings, Alan reasoned, it might be worth educating the modules— investing resources in making each module smarter and more flexible—and then relying on the resultant community of agents to solve problems in ways not fully preconceived by their developers. In other words: make them smarter, better informed, give them higher level responsibilities, and then give them more freedom to meet their goals. This zeitgeist thundered through Carl Hewitt’s ACTORS [43], Seymour Papert ’s LOGO [51], my own BEINGS [46], and of course Alan Kay and Adele Goldberg’s Smalltalk [40], and has reverberated ever since, sometimes explicitly as in the case of Mixins (in Lisp Machines FLAVORS [52] and, more recently, in Java) and sometimes almost unrecognized as in the case of Expert Systems [42] and, more recently, the Semantic Web.