The cycle that began with [YOU] on AI celebrates the twenty-fold productivity multiplier achievable when a single engineer in Trivandrum works with Claude instead of working in a team of twenty. Wenger's framework accepts the celebration and immediately asks the harder question: what kind of knowledge did those twenty engineers produce that the solo builder does not? The answer is not a matter of quantity but of kind. The team was a community of practice—a group bound by shared domain, mutual engagement, and a collectively maintained repertoire of stories, standards, and sensibilities that no training manual can encode and no AI can replicate, because it lives not in any individual or any document but in the ongoing interactions among practitioners who share stakes in the work.
Wenger's 2023 analysis of generative AI, published with colleagues, deploys this vocabulary with precision that no productivity-focused framework can match. AI systems, he argues, are reifications—sophisticated artifacts that encode patterns from vast quantities of human-generated text. Reifications can be extraordinarily useful. But reification without participation is dead: it has the form of knowledge without the lived experience of generating, contesting, and transmitting it. AI as boundary object describes this role precisely. Claude bridges between the designer's vocabulary and the engineer's, between the analyst's priorities and the developer's task list, with remarkable fidelity—but a boundary object, in Wenger's framework, is not a community member. It does not hold its interlocutors accountable to shared standards. It does not generate the joint enterprise through which communities define themselves.
The dissolution Wenger's framework predicts is not sudden. Communities of practice erode. When the junior developer's questions go to Claude at 3 a.m. rather than to a senior colleague over coffee, the colleague does not receive the question. The relationship does not deepen. The community's shared repertoire does not accumulate one more collectively processed problem. Legitimate peripheral participation—the structured entry point by which newcomers absorb the community's tacit knowledge through immersion in peripheral tasks—is disrupted at its foundation when the peripheral tasks are automated. The garment is produced. The practitioner is not.
His lens reframes the deskilling debate that runs through the cycle. Deskilling is the wrong word for what Wenger identifies. The issue is not that capabilities are lost but that formation is interrupted. The engineer who produces senior-level output on day one has the capability without the formation—the judgment, the professional identity, the somatic knowledge of what production failures feel like under pressure—that legitimate peripheral participation would have deposited. The capability and the formation are both real; both matter; they are produced by different processes; and AI augments the first while structurally reducing the second.
Born in Switzerland in 1952, Wenger trained as a computer scientist before moving into educational research, a trajectory that gave him both the technical literacy to understand what AI tutoring systems could do and the disciplinary distance to see what they assumed. His encounter with Jean Lave at the Institute for Research on Learning in Palo Alto in the late 1980s was the decisive event. Lave had spent years embedded in non-Western apprenticeship settings—tailors in Liberia, midwives in the Yucatan, butchers in American supermarkets—and had reached a conclusion that cut against everything the AI-in-education field presupposed: learning was not the internalization of knowledge by an individual mind but the transformation of participation in a social practice. The Liberian tailor's apprentice did not learn by receiving instructions. She learned by participating in the community's practice, starting at the periphery and moving gradually toward full membership.
The collaboration produced Situated Learning (1991), with its central concept of legitimate peripheral participation. Wenger's solo work, Communities of Practice (1998), built the full theoretical architecture: the three-element definition of a community of practice (shared domain, mutual engagement, joint practice), the dual processes of participation and reification through which communities produce meaning, and the account of identity as constituted through practice rather than as a psychological given that precedes it. His later work turned prescriptive, addressing how organizations can deliberately cultivate the conditions under which communities of practice form and thrive.
The man who had written the definitive survey of AI tutoring systems concluded that the entire paradigm rested on the wrong model of learning. The conclusion cost him nothing professionally—the framework he built in its place became one of the most influential in educational research, management theory, and organizational learning. But it carries a irony that the AI moment amplifies: the researcher who most clearly diagnosed the limits of machine-mediated learning in 1987 has produced the framework that most precisely diagnoses the limits of the far more sophisticated machine-mediated learning that arrived in 2025.
Community of Practice. Wenger's foundational unit is not the individual learner or the organization but the community of practice: a group bound by shared domain (members care about the same thing and recognize each other as fellow practitioners), mutual engagement (sustained interaction through which relationships, reputations, and trust develop), and a shared repertoire (the accumulated stories, tools, procedures, and sensibilities that the community has produced through joint work). All three elements are necessary; none is sufficient alone. The community of practice is where the deepest professional knowledge lives, and it is what the solo AI builder structurally cannot replicate.
Participation and Reification. Communities produce meaning through two complementary processes. Participation is the direct, lived experience of engaging in a practice—the conversations over coffee, the code review that catches a pattern of thinking rather than a bug, the architectural debate where two perspectives collide. Reification is the giving of form to experience—documents, tools, procedures, concepts. The interplay is essential: reification without participation is dead form, accurate in content but disconnected from the lived experience that gives it meaning. AI systems are reifications. The most sophisticated boundary object in the organizational world is still a reification—and a reification cannot substitute for participation.
Legitimate Peripheral Participation. Newcomers become practitioners not by receiving knowledge but by participating in the community's practice from the periphery, moving gradually toward full membership as their competence develops and the community recognizes it. The periphery is a learning position, not a position of exclusion: peripheral tasks are real, they contribute real value, and immersion in them deposits the tacit knowledge that no explicit instruction can convey. When AI automates the peripheral tasks, the entry point into the community's formation disappears. The output continues; the formation does not.
Identity Through Practice. In Wenger's framework, professional identity is not a psychological attribute preceding practice but is constituted through practice. To become a software developer is not merely to acquire programming skills; it is to become someone whose sense of self is organized by membership in a community of practitioners who share a domain, a repertoire, and a set of standards for what good work looks like. The identity crisis the cycle documents—the senior engineer in Trivandrum who spent two days oscillating between excitement and terror—is, in Wenger's terms, the instability of an identity constituted through a practice that AI has transformed.
The Constellations of Practice. Beyond individual communities, Wenger identifies constellations of practice—networks of related communities held together by shared histories, overlapping repertoires, and the artifacts that travel between them. AI does not merely disrupt individual communities; it reorganizes the constellations, creating new affiliations between builders who share only a tool and dissolving older affiliations built around shared craft. The organizational consequences extend far beyond the productivity gains that any single builder-AI interaction can measure.