Autopoesis is both a concept about self-maintaining systems, especially biological ones, and an epistemological theory about cognition and self-reference. 

Humberto Maturana and F. J. Varela proposed the concept of autopoesis (as opposed to allopoetic) as the organization of living things, which maintains their lasting unity and wholeness in the midst of continual turnover of constituents, specifically through self-production. For them, autopoesis is the necessary and sufficient condition for the minimal units of life. Autopoesis defines the minimal mechanistic features of the simplest possible living entity. It describes "how life works." All living systems are autopoetic, and all physical systems, if autopoetic, can be said to be living. 

An autopoetic system is defined as an entity that can be clearly distinguished and is realized through a closed organization of production processes such that the products of the process interact in the same way as the organization that produces them and a discernable topological boundary emerges as a result of these same processes. ( Artificial Life, p.131) An autopoetic unit is a system that makes itself, through a network of interactions that take place within its own well-defined boundaries. (Cf. autocatalysis) Or, as Maturana puts it in Autopoesis and Cognition, "The living organization is a circular organization which secures the production or maintenance of the components that specify it in such a manner that the product of their functioning is the very same organization that produces them." (p.48) The goal of autopoetic systems is the maintenance of their internal organization. Thus self-completion or individuation is distinctive to the living state. The capacity of active self-maintenance and self-generation that underlies regeneration, reproduction, and healing, are ways that organisms become distinctive wholes. The autopoetic system is not entirely autonomous, however. It must be structurally coupled to to the world, to the medium in which it exists.

According to Katherine Hayles, the "second wave" of cybernetics started with an article entitled "What the Frog's Eye Tells the Frog's Brain," in which the authors, including Humberto Maturana, showed that a frog's visual system does not so much represent reality as construct it. As the authors noted, "The frog's eye speaks to the brain in a language already highly organized and interpreted instead of transmitting some more or less accurate copy of the distribution of light upon the receptors." (How We Became Posthuman, p.131, p. 135) (see vision ) Humberto Maturana developed the epistemological implications of this study by stressing the internal organization and circularity of cognition and of living systems in general. It is this recursivity which characterizes "second order" cybernetics, which entails the observation of observation (in a formal, and not psychological frame.) Thus, second-order cybernetic systems are both closed and open. Autopoetic systems are closed at the level of operation or organization, but are open at the level of structure.

For Varela, "An autopoetic machine continuously generates and specifies its own organization through its operation as a system of production of its own components." (cf. Kant's distinctions between mechanism and organism in mechanism / vitalism) Note, however, that the theory of autopoesis defines the relations of the components without reference to the whole. Félix Guattari argues for a machinistic autopoesis, "the non-discursive and self-enunciating nexus of the machine." (as a critique of any unifying idea of the signifier -- esp. the Lacanian one)This autopoetic nexus "wrests it from structure."He cites Varela's description of a machine as "the ensemble of the interrelation of its components, independent of the components themselves". Despite the fact that Varela reserves the qualification of "autopoetic" to the biological domain, Guattari extends the notion accross the range of social and technical processes that he includes in "machinism." Guattari is not so interested in autopoesis as unitary individuation but rather in entities that are "evolutive and collective." According to Guattari, "We can thus envisage autopoesis under the heading of an ontogenesis and phylogenesis specific to the mecanosphere that superimposes itself on the biosphere." 

Following the distiction established by Maturana and Varela in Autopoesis and Cognition, Guattari distinguishes allopoetic machines from autopoetic machines. The former produce something beside themselves, and perform a function useful to agents outside themselves, whereas the latter continuously engender and specify their own organization and their own limits. ("Machinic Heterogenesis" in Rethinking Technologies, p. 17) He also connects the idea of the allopoetic machine with hypertext. ("On Machines" in JPVA no.6) But if Varela's emphasis is on autopoesis as unitary individuation, Guattari stresses a more collective machinism, a "diagrammatic rhizome" of ontological heterogenesis. (see machinic phylum)

see also network or assemblage 
Thus the basic autopoetic unit is closely linked to the cell, defined by its boundary, with a metabolic network inside. It is self-bounded, self-generating, and self-perpetuating. The potential for self-reproduction and evolution are consequences of the initial definition of autopoesis. According to Pier Luigi Luisi, the minimal autopoetic unit requires a process of self-generation and a process of decomposition. If the rates of the two process are equal, the system will self-perpetuate in a steady (homeostatic) state. If the rate of the generative process is greater, the unit will tend to grow, and if there are limits to its size, it will tend to self-replicate. In the other case, where the rate of decomposition is greater, it will self-implode. These are the three basic kinetic states. 

Luisi claims that according to this definition, viruses are non-living. They are certainly not free-living. They are parasites that invade cells, co-opt the cell's metabolic machinery to accomplish their own self-reproduction, escape the host cell, and invade another... nor is a robot that assembles other robots out of parts built by other machines, because in both cases reproduction is not occuring within the organizational domain of the unit.

Autopoetic systems have the ability to discriminate what is part of the system and what is not, which provides a link to a theory of observation. Is an autopoetic system a self-describing system? 

When brought to bear on theory and criticism, autopoesis establishes a new epistemological condition. The the observer is not separate from what is being observed, but is structurally coupled to the phenomenon she sees. Stating that "everything said is said by an observer," Maturana and Varela develop a theory of the observation of observation, in which the world is not given -- as in the traditional representationalist frame -- but is rather brought forth in the dynamic interaction of observer and observed. "Our intention is to bypass entirely this logical geography of inner versus outer by studying cognition not as recovery or projection but as embodied action." (Tree of Knowledge, p. 172) Autopoesis leads to a theory of observor in which there is no route back from the act of observing to the data that was used to generate the theory in the first place. (Hayles) 

Niklas Luhmann describes complex self-referential systems as "hypercomplex," in the ways that they make distinctions between themselves and their environments, but also internalize the distinction in a paradoxical "re-entry" of the distinction into the distinguished. 

Autopoesis: (See The Tree of Knowledge, by H.R. Mataruna and F. G. Varela. and also "Autopoesis: the Organization of Living Systems, its Characterization and a Model" Biosystems 5 . See also Francisco Varela, Autonomie et Connaissance), Pier Luigi Luisi, "Defining the Transition to Life: Self-replicating Bounded Structures and Chemical Autopoeisis", in Stein & Varela, eds. Thinking About Biology.