Assemblage

In the AI Story

Deleuze and Guattari developed the assemblage concept partly in response to structuralist and systems-theoretical approaches that tended to treat social formations as closed totalities with their own laws of operation. The assemblage is not a closed system; it is open, temporary, composed through specific connections that can be reconfigured. It has an inside only in the sense that its current connections distinguish it from other possible configurations, not in the sense that it has a fixed boundary. This openness is what makes the assemblage a useful tool for analyzing human-machine relationships: it refuses the fantasy of a pure human who stands before a separate tool, and it refuses the opposite fantasy of a totalized system that absorbs the human completely.

In the AI context, the developer-AI assemblage is a concrete instance of the concept. When a developer works with Claude Code, the resulting output — a piece of software, a body of documentation, a design — is not attributable to either the human or the machine alone. It emerges from their interaction: from the developer's intentions, tacit knowledge, and aesthetic judgments combined with the model's training data, statistical patterns, and response generation. The code that results cannot be cleanly separated into human-written and machine-written portions; it is a product of the assemblage.

This framing has consequences for the Orange Pill's central question. Are you worth amplifying? presumes a coherent you that exists prior to and independently of the amplifier. The assemblage framework suggests that this presumption is exactly what the AI relationship makes untenable. The developer who has worked with AI for months — whose creative habits, problem-solving strategies, and sense of the possible have been shaped by continuous interaction with the tool — is not the same developer who began the first session. The tool has not merely amplified a pre-existing self; it has participated in the production of a new self, a new assemblage that is neither purely human nor purely technological.

The assemblage concept also illuminates what can and cannot be said about responsibility in AI-assisted work. Traditional ethics presumes a human agent whose choices can be evaluated against standards of right and wrong. The assemblage framework complicates this evaluation: if the output emerges from the interaction, who is responsible for it? The question is not evaded by the framework but reframed. Responsibility attaches to the configuration of the assemblage — to how the elements are combined, what flows between them, which capacities are developed and which are atrophied through the relationship. The ethical question is not merely about individual choices but about the specific mode of combination through which human and machine work together.

Origin

The concept of agencement was developed primarily in A Thousand Plateaus (1980), building on earlier work in Anti-Oedipus and on Foucault's analyses of discursive formations. The English translation assemblage has been criticized for missing the French connotations of arrangement, connection, or agency (from agir, to act), but has become standard in Anglophone scholarship. Manuel DeLanda's A New Philosophy of Society (2006) offered an influential extension of the concept into social theory, though some Deleuzians have argued that DeLanda's version loses the political edge of the original.

Key Ideas

Assemblages are heterogeneous. They combine elements of different ontological status — human, technological, institutional, semiotic — without reducing them to a common substance.

Assemblages are emergent. Their effects cannot be predicted from the properties of their components; they emerge from the specific ways elements are combined.

Assemblages are temporary. They are held together by connections that can be reconfigured, dissolved, or extended into new configurations.

The human is always an assemblage. There is no pure human prior to tools, institutions, and social relations; the human has always been a specific mode of combination.

The developer-AI relationship is an assemblage. What emerges from AI-augmented work cannot be cleanly parsed into human and machine contributions; it is a product of a specific configuration that shapes both components.

Debates & Critiques

The assemblage concept has been extended, contested, and sometimes watered down in subsequent theoretical work. Actor-network theory (Bruno Latour and others) developed related concepts with somewhat different implications. Some critics have argued that assemblage thinking dilutes political critique by dispersing agency across networks in ways that make responsibility hard to assign; defenders argue that the concept reframes rather than eliminates responsibility, asking how assemblages can be intentionally composed to produce desirable effects rather than left to form through default dynamics. The Deleuze volume takes the position that the assemblage concept is indispensable for thinking about AI because it dissolves exactly the boundaries — between human and tool, between agency and structure — that the AI moment has made analytically untenable.