Manuel Castells
Power is not about who sits at the top of the organizational chart. It is about who is most connected in the network. Castells stated this in 1996. Amazon, Google, and Meta confirmed it later.
The rise of the network society
Article 18 of 21—Governance as Versioning Series
This is Article 18. The Governance as Versioning series maps 335 years of political, philosophical, and systemic thought to the logic of code versioning. Article 18 covers Manuel Castells, the transversal concept of topology as power—the idea that each participant’s position in the network of contributions, reviews, and approvals is itself a form of political influence—and the question that connects Lessig to Schulz: if the network’s structure is the structure of power, then the most important political act is not who gives orders, but who designs the connections.
What you’ll find here:
This article shows that power is not who sits at the top of the org chart: it is who is most connected in the network. This discovery transforms how we read a collective versioning system. Visible contribution, traceable history, and position in the review network are the new forms of influence, with or without a title. The symmetrical risk is also here: a network can exclude without anyone planning to exclude—simply through the invisible design of whoever built it.
1. Where it all begins
Barcelona, 1966. Manuel Castells is 24 years old and has just been expelled from Spain by Franco’s regime. The crime? Participating in a student strike. He flees to Paris with no money, no job, and no plan. He gets a scholarship, earns a doctorate in sociology, and spends the following years trying to understand a question that will not leave his mind: why does power sometimes sit where it appears to sit and sometimes vanish, invisible, to somewhere no one can point to?
Three decades later, in Berkeley, California, he finally has the answer. And it is in the most obvious and most overlooked place at the same time: in the structure of connections.
It is 1996. The commercial internet is only two years old. Most academics still think it is a passing curiosity, a thing for young enthusiasts. Castells, sitting at the heart of Silicon Valley, sees something else. He sees that this is not about technology. It never was. It is about how power changes hands when the way of connecting people changes form.
He publishes the first volume of a trilogy that will run to more than 1,500 pages: The Rise of the Network Society. The central argument fits in one sentence: the basic unit of social organization is no longer the individual, the family, the company, or the state. It is the network.
It seems obvious today. In 1996, it was radical. Almost no one understood what it really meant, nor the consequences it would come to have.
2. Historical context
Castells writes at a moment of transition that no one knows how to name. The Cold War ended in 1991. But the old industrial structures are in crisis: factories closing, jobs vanishing, and unions losing strength. At the same time, companies without physical assets appear in California, are worth billions, and no one quite knows what they produce.
The dominant social theory still uses nineteenth-century categories: classes, states, and hierarchies. It works for 1900. It does not work for what is being born. Castells realizes a new category is needed. The main one is the network.
3. The central idea
Before Castells, the question about power was, who is at the top? King, CEO, president, director-general. Power descended from above. The org chart was the real constitution of any institution.
Castells turned that question upside down. The right question is not who is at the top. It is: who is most connected?
Think of a city before and after the subway. Before, the most powerful neighborhood was the one closest to the center. After, the most powerful is the one with the most lines passing through it. The center may even shift, because power now depends on position in the network, not on the address on the map.
The same happens with organizations. In a hierarchy, power is position: at the top, you give orders. In a network, power is centrality: the more connections you have, the more information passes through you, and the more influence you exercise, without needing to give a single order.
This has three consequences that Castells works through carefully.
First: whoever is outside the network simply does not exist for the system. They are not marginalized, not oppressed, not exploited. They are invisible. As if they were not there. This is a new form of exclusion, more silent and more permanent than the ones before it.
Second: networks are faster, more adaptable, and more resilient than hierarchies. Hierarchy is efficient when the environment is predictable. Networks are superior when the environment changes constantly. That is why social movements without central leadership survive repression: cut off one head, the network reorganizes. That is why networked companies innovate faster than bureaucracies.
Third, and most important: power in a network is not exercised by command, but by architecture. Whoever designs the network defines what is possible, what is easy, and what is blocked—without any of it appearing in any manual.
4. Developments
Space of flows: the power that lives nowhere
Castells coins a concept that seems technical but is disturbing: the space of flows.
Most of us live in the space of places. We live in neighborhoods, work in specific buildings, and vote in cities. Our lives have an address.
Power, increasingly, operates in the space of flows. Financial capital moves between continents in milliseconds. A company registered in the Cayman Islands, with servers in Ireland, operations in Brazil, founders in California, and customers everywhere. Where does it exist? Nowhere fixed. It exists in the flows between those points.
This creates a political problem that remains unsolved. Elected politicians control territories. The power that affects people’s lives operates in flows that ignore borders. The result is that familiar discomfort: the sense that voting changes less than it should.
Castells is not complaining. He is describing a structural shift. But describing it precisely is the first step toward dealing with it.
The power to program: who writes the network’s rules
Castells identifies three forms of power in networks. The most important, and the most invisible, is the power to program: whoever defines the operating rules of the network controls what is possible within it.
It is not censorship. It is design.
Facebook does not prohibit you from posting. But the algorithm decides who sees what you posted. If it does not distribute, it is as if you had not posted. Power does not lie in prohibition: it lies in the architecture of visibility.
Google does not delete websites. But the ranking decides whether anyone will find them. Page 10 is invisibility.
Castells described this in 1996, before social networks existed. He was articulating a general principle: in networks, whoever writes the protocol holds political power, without election, without accountability. Lessig, in the previous article, showed that code is law. Castells completes the argument: network is power.
5. Historical legacy
Castells’s trilogy sold more than a million copies and was translated into 23 languages. The most concrete legacy, however, lies in what the world confirmed afterward.
The Arab Spring, in 2011, toppled decades-old dictatorships with movements that had no central leadership. Castells analyzed the phenomenon in Networks of Outrage and Hope (2012): the form of organization mattered as much as the content of the demands. The 2013 protests in Brazil repeated the pattern: called without formal hierarchy, amplified by networks.
Get Gustavo J. Alessandri’s stories in your inbox
Join Medium for free to get updates from this writer.
Amazon, Google, and Meta confirmed the central prediction: the power of the twenty-first century belongs to whoever controls the platforms where economic and social life happens.
6. Contradictions and limitations
Castells has an important blind spot: he tends to see the network as a structure that can be used for good or for ill. But networks are not neutral. They are designed by someone, in some context, with certain interests. The network’s design is already a political choice.
Furthermore, Castells’s analysis is fundamentally descriptive. He is extraordinarily precise in diagnosing what is happening. But on what to do about it, the answers turn vague. “Redesign institutions” is advice that does not resolve the problem of who has the power to redesign them and through what process.
This teaches us that even the best structural analyses need to be complemented by an analysis of agency: who are the concrete people who can change something, and how do they build the power to do it?
7. The translation into versioning
A legitimate objection before continuing: “If power in a network belongs to whoever is most connected, then a collective versioning system will naturally concentrate influence in the most active participants, reproducing a new informal hierarchy. How is that different from having a boss?” The answer lies in visibility: in a traditional hierarchy, influence comes from a title and need not be demonstrated. In an open repository, influence comes from the public history of contributions, reviews, and discussions—which any participant can verify, contest, and eventually surpass. It does not eliminate inequality of participation, but it makes that inequality auditable.
Castells never used Git. But had he, he would have found the best practical implementation of everything he theorized.
The transversal concept Castells illuminates in this series is topology as power: the idea that each participant’s position in the network of contributions, reviews, and approvals is itself a form of political influence. No single step of the protocol captures this, because topology emerges from the whole: each commit (a recorded change), each review (an assessment), and each approval or rejection builds or dismantles a participant’s position in the network.
Power as a verifiable trajectory, not as a title. In a repository, who opens the most issues, who reviews most consistently, and who proposes merges that get approved? This is the public history of contribution. It is not a title that someone grants. It is a trajectory that anyone can read, compare, and question. Influence exists, but it is exposed.
Invisible exclusion as a design risk. An open repository is not the same as an accessible repository. If the documentation is only in English, if the contribution process requires technical familiarity that most people lack, and if discussions happen in jargon that excludes the formal openness, then the openness is illusory. The network exists. But some people are outside it, invisible to the system, without anyone having planned that.
The power to program the repository’s own rules. Whoever sets the permissions, whoever configures what a review must include to be valid, and whoever decides the structure of categories and progress milestones—that person is exercising the system’s most important power. Not by command, but by architecture. That is why the process of defining the network’s rules must itself follow the rules it espouses: openness, record-keeping, and collective review.
Vigilance: an open governance system can close without anyone noticing, simply through the accumulation of architecture decisions that progressively narrow who manages to participate. It is not censorship. It is design. And invisible design is the most dangerous design.
A limitation versioning does not resolve on its own: the protocol assumes good faith from whoever configures the system’s initial topology—the permissions, the roles, and the approval criteria. Nothing prevents someone from building a formally open but structurally concentrated network, where the central nodes belong to a restricted group and other participants exist at the periphery without real influence. What open history offers is the possibility of auditing that concentration and naming it. Visibility does not solve the problem, but it is the condition for the problem to be solvable.
8. Practical implication
OpenStreetMap in Brazil: when the network reveals what the map hides
OpenStreetMap (OSM) is a worldwide map built by volunteers, with no controlling company, no formal production hierarchy. Anyone with internet access can register a street, correct the name of a square, and mark a hospital. The result belongs to everyone, under an open license. In 2024, the project had more than ten million registered contributors worldwide, and the map it generated was used by humanitarian organizations, researchers, and logistics companies in dozens of countries.
In Brazil, the contribution pattern illustrates Castells’s thesis about invisible exclusion with precision. High-income neighborhoods in São Paulo, Rio de Janeiro, and Brasília were mapped in detail comparable to European cities: streets, establishments, sidewalks, and even trash cans. Peripheries of those same cities, favelas, and interior municipalities appeared as white patches or with decades-old information. Not because anyone had prohibited mapping those areas. But because the contributor network was topologically concentrated, those who had time, a device, and the technical familiarity to contribute lived predominantly in higher-income urban centers.
The exclusion was not intentional. It was architectural. The network was open. But the distribution of who managed to enter it reproduced the inequalities of the physical world.
The answer came through redesigning the network, not through decree. Humanitarian mapping projects, like the Humanitarian OpenStreetMap Team (HOT), created simpler contribution tools, trained local communities, and organized mapping sprints in underrepresented areas. The change in the participation architecture altered the network’s topology. Castells had described the mechanism thirty years earlier.
Now imagine this: a craftworkers’ cooperative tried to create a shared repository of models and designs. They published the link and wrote the rules. Three months later, seventeen people had access. Four were contributing. Thirteen had never logged in.
The problem: instructions in technical language, a process that required interface familiarity that not everyone had, and reviews scheduled at a time that excluded those working the afternoon shift. The network existed. But it was programmed to exclude it, without anyone having planned it.
The solution came through redesigning the architecture: plain-language instructions, contribution by form, and asynchronous reviews with flexible deadlines. Regular contributors tripled in two months. Castells had already described the mechanism. They had to discover it for themselves.
9. Lessons learned
Formal openness is not real openness: the network needs to be designed to include, not merely to permit others to enter if they can manage it.
Power in networks is topological: it lies in position and connections, not in title. It can be more just than hierarchy, or it can replicate old inequalities in new clothing.
Network design is a political decision, especially when it appears technical.
Whoever is outside the network is invisible to the system. This exclusion is harder to fight because it leaves no trace of intent.
Redesigning the participation architecture is more effective than publishing inclusion rules. Rules say what should happen. Architecture defines what actually happens.
10. Exploring paths and connections
Eighteen articles completed. Each thinker resolved one problem and created another: Locke gave consent; Popper showed that consent without error correction is not enough; Hayek showed that correction requires dispersed information. Mises showed that order can emerge without a planner; Shannon measured information; Dawkins showed how ideas survive; Johnson saw emergence in simple rules; Weber warned about the cage; Lippmann about the distorted images; Kahneman about the mind’s shortcuts; Wurman about the excess that paralyzes; Ostrom about the commons; Meadows about the feedback loops; Dewey about democratic learning; Surowiecki about the crowds that get things right; Benkler about the peers who produce; Lessig is about the code that governs. Castells closes the arc: the structure of the network is the structure of power.
But all of this system, however well designed, passes through people. And people are wrong. Not as an exception: as a rule.
It is 2010. An American philosopher named Kathryn Schulz is finishing a book about something everyone experiences and no one wants to admit. She will show that the problem is not the error itself. It is that being wrong feels exactly like being right—until the moment you discover you were not. Article 19 begins exactly there.
Previous articles, the preface, the introduction, and full documentation are available here:
https://codeberg.org/openarchy/gitgov
https://github.com/openarchy/gitgov
All content is fully open source and accessible to everyone. Licensed under CC BY-SA 4.0: read, share, translate, adapt, and improve—as long as you keep attribution and distribute derivatives under the same license.
If you find an error, have an idea, or want to propose an improvement, just open an issue or fork it on Codeberg or GitHub. Contributions are welcome. That’s exactly the point.
11. References
- CASTELLS, Manuel. The Rise of the Network Society. Oxford: Blackwell, 1996.
- CASTELLS, Manuel. Networks of Outrage and Hope: Social Movements in the Internet Age. Cambridge: Polity Press, 2012.
- CASTELLS, Manuel. Communication Power. Oxford: Oxford University Press, 2009.
- HUMANITARIAN OPENSTREETMAP TEAM. HOT — Humanitarian OpenStreetMap Team: about. Washington, D.C.: HOT, 2024. Available at: https://www.hotosm.org Accessed: Mar. 2025.
- OPENSTREETMAP FOUNDATION. OpenStreetMap: the free wiki world map. London: OSMF, 2024. Available at: https://www.openstreetmap.org information. Accessed: Mar. 2025.
