Architects are creative problem solvers, primarily driven by intuition, while coming from a sense of the past and the logic of the present. Our initial attempts at integrating computing into the studio, as evidenced by this collection of papers, is very diverse, based on differing pedagogical assumptions, and the achieving of significantly different results. This would appear to be evidence of a revolutionary approach to the problem rather than a scientific evolutionary approach. Terrific! This is when we as architects are at our best. Although we reach a great number of emphatically dead ends, the successes and discoveries achieved along the way are significant.

The diversity and quality of papers submitted suggest that we are indeed pursuing the task of integration in our typical, individual, intuitive, logical manner. I commend all of the authors who submitted proposals and thank them for expanding the envelope of integration into their personal exploration.

Computer programs exist that are well suited for this purpose, two of which are described. They offer a broad range of design capabilities, and are easy to master because of their intuitive and graphically oriented approach.

A number of examples are given to illustrate the potential of computer-aided learning as a complement to traditional methods either in the classroom or in coursework.

A highly unusual feature of PHIDIAS II is that it implements all of its functions using only hypermedia mechanisms. Complex vector graphic drawings and objects are represented as composite hypermedia nodes. Inference and critiquing are implemented through use of what are known as virtual structures [Halasz 1988], including virtual links and virtual nodes. These nodes and links are dynamic (computed) rather than static (constant). They are defined as expressions in the same language used for queries and are computed at display time. The implementation of different kinds of functions using a common set of mechanisms makes it easy to use them in combination, thus further augmenting the system's functionality.

PHIDIAS supports design by informing architects as they develop a solution's form. The idea is thus not to make the design process faster or cheaper but rather to improve the quality of the things designed. We believe that architects can create better buildings for their users if they have better information. This includes information about buildings of given types, user populations, historical and modern precedents, local site and climate conditions, the urban and natural context and its historical development, as well as local, state and federal regulations.

The basic idea is that of top-down design - beginning with a very abstract representation, and elaborating that, in step-by-step fashion, into a complete and detailed representation. The basic operations are real-time parametric variation of designs (using the mouse and slide bar) and substitution of objects. Essentially, then, a knowledge-base in Topdown implements a kind of parametric shape grammar.

The main applications of Topdown are in introductory teaching of CAD, and (since it provides a very quick and easy way for a user to develop detailed geometric models) to provide a uniform front-end for a variety of different applications. The shell, and some example knowledge-bases, are publicly available.

This paper discusses the principles of the Topdown Shell, the implementation of knowledge bases within it, and a variety of practical design applications.