No community of practice exists alone. Organizations of any complexity are constellations — networks of related communities connected by shared members, boundary objects, histories, and concerns. The frontend team, the backend team, the design team, the product team: each is a community of practice; together they form a constellation whose health depends on the density and quality of connections across it. Wenger's framework reveals that the most significant learning often occurs not within communities but at the boundaries between them — at the friction points where different practices collide and generate insights neither community would produce alone. The AI age, by dissolving the connective tissue between communities even as it improves the coordination of their outputs, threatens constellations with a specific pathology: more integrated products, less integrated practice.
The constellation is held together by four primary mechanisms: shared members (practitioners who belong to multiple communities), boundary objects (artifacts usable across community boundaries), boundary encounters (direct confrontations between communities), and shared histories (accumulated collective experience that spans communities). AI disrupts different mechanisms differently — it multiplies boundary objects while reducing shared memberships and boundary encounters.
The connective tissue of a constellation is often invisible to formal organizational analysis. It runs through overlapping meetings, informal conversations, shared projects that cross team lines, the careers of practitioners who move between communities. When this tissue is cultivated, the constellation generates cross-community learning. When it thins, each community becomes more self-contained — more expert in its own practice, less able to see how its practice relates to others.
The AI paradox for constellations is that outputs become more integrated while practices become more siloed. Claude translates perfectly between the designer's vocabulary and the engineer's; the product is seamless. But the designer and the engineer no longer meet at the boundary, and the cross-community understanding that boundary encounters produced does not develop. The next generation of practitioners inhabits a world of integrated products and siloed practices — and may never develop the constellation-level perspective that was once a natural outcome of working across boundaries.
The emerging constellations forming around AI — the builders posting techniques, the conferences on AI-augmented development, the online discourse that The Orange Pill describes as 'the look of recognition' between practitioners who have crossed the threshold — represent constellation formation in progress. Whether they develop the depth required to substitute for the constellations they are replacing is the open question, with implications for a generation of professional learning.
Wenger developed the constellation concept in Communities of Practice (1998) to address a limitation of the original framework — its focus on individual communities obscured the interconnections through which organizations actually function. The extension became especially important in his later work on cross-organizational learning networks, where the constellation level was where much of the most consequential work happened.
The concept has resonated with research on knowledge ecosystems, innovation networks, and inter-organizational collaboration, all of which suggest that the most productive configurations for complex knowledge work involve multiple communities in ongoing interaction rather than either isolated communities or a single undifferentiated network.
Networks of communities. Not a single community and not an undifferentiated network — communities in ongoing interaction.
Held together by connective tissue. Shared members, boundary objects, encounters, histories.
Boundaries are sites of learning. The friction between different communities generates insights neither would produce alone.
AI paradox. More integrated outputs, less integrated practice — a specific failure mode the constellation framework makes visible.
New constellations are forming. Around AI-augmented building, AI governance, domain-specific AI use — their depth will shape professional learning for a generation.
Whether the new constellations forming around AI can achieve the depth of the old ones they are replacing is contested. Some argue that digital-native communities will develop analogous depth through different mechanisms (asynchronous communication, global scale, rapid iteration); others argue that the conditions for deep community formation — sustained mutual engagement, identity-shaping trajectories, developed shared repertoires — require the time and proximity that AI-augmented work often eliminates.