The following paper attempts a reflection on interwoven notions like ‘system’, ‘evolution’ and ‘design’. To be interwoven these notions must be observed as threads - and they have all been theory threads from the nineteenth century to the present - and lines can be confused with edges, cuts, fractures, borders or distinctions. Someone is expected to draw these distinctions and someone (maybe, but not necessarily, someone else) must be addressed by them; the notions must be ascribed in both ways. Thus, ‘system’, ‘evolution’ and ‘design’ are conditioned by the notion of ‘observer’ and indicate nothing other than observation processes. Niklas Luhmann was always in doubt as to whether such notions could orient scientific research in any traditional sense; in 1987 he pointed out that the question would be if and how an observation-based society like ours could ever have the ability to develop a stability of structures, when all of its forms are observed as social systems being observed and as observing systems. There are, he noticed, only rats in mazes, who observe each other, and a theory describing this situation certainly could be called a rat theory.

Heinz von Foerster described such a rat theory as a theory of ‘observing systems’ and so he developed the strand of systems theory and second order cybernetics which I am interested in here. The most important thesis of such theories is that the one who draws a distinction (the observer) has to take into account that he or she is always included in his or her distinction. There is no sovereign external point of view to look at the situation, or in other words: the only sovereignty or the only playground that an observer can occupy is inside, not outside his or her observation: ‘Tertium datur’. That's the crucial point that couples system, evolution and design. It is no longer necessary then to describe learning as a search for something that is always already known; it is observed here as a switch between a goal strategy (which refers to knowledge) and an exploration strategy (which refers to non-knowledge). The exploration strategy is subversive because it does nothing other than undermine the goal strategy, including former knowledge by forgetting it. My paper discusses these intertwined inclusions (the inclusion of the observer and the observed into the observation and the inclusion of knowing and not-knowing into learning) using three of Claude Elwood Shannon's amazing machines as examples: the maze-solving machine, the four-dimensional turning machine for American car-drivers in English traffic, and the ultimate machine, which works by switching itself off. All of these machines couple their