Join Discord
Modular and Portable Multiplayer Miniverses

Modular and Portable Multiplayer Miniverses


A growing number of subcultures, digital communities and guilds have turned their back on ad-supported social media and migrated their social and cultural activities to semi-private digital spaces, chat rooms and Discord servers. We believe these spaces have the potential to become decentralised institutions that are financed, owned and governed by their own members. To support this vision we propose Moving Castles, an organisational metaphor and real-time media type which combines collective agency and public participation in modular and portable multiplayer miniverses.

A Moving Castle crosses the ridge between private and public.
A Moving Castle crosses the ridge between private and public.

A Moving Castle—as popularised by Studio Ghibli’s cult movie “Howl’s Moving Castle”, based on Diana Wynne Jones’ 1986 book of the same title—is a nomadic patchwork of different spaces of mutating and incoherent scale. Such spaces are “bespelled to hold together’’ by the will of their inhabitants, which the Moving Castle can transport to different worlds in times of needs or wish.

“With a last squeak the castle lifted, the crew cheered as we could see across the dark forest for the first time, our eyes locked on the arid wastelands on the horizon. Now back to chat, this machine won’t move itself”

In our vision of this new media format, Moving Castles are modular and portable multiplayer miniverses; inhabited by communities that use them to manage their lore, ecosystems and economies. With this first post, we sketch their blueprints and collage the new collectively produced media-format that can emerge from within Moving Castles. After collecting the seeds from which these communities could grow and sustain themselves, we’ll finally take a look at how they can be built and how to facilitate an exchange with a public outside their familiar territories.

Not islands or dark forests, but vehicles

We are part of Trust, a knowledge-community that lives on a member-restricted Discord. The focus over the last year has been on producing online events which take shape as Collective Reading Groups, Sandbox (a combined voice and chat presentation format) and publicly available streams and talks on Twitch.

Trust and a large variety of other semi-private communities—from pay-to-enter-discords, game guilds, blockchain communities, subscription-mediated newsletters, and group chats— are what Yancey Strickler calls Dark Forests. According to him these are “spaces where depressurized conversation is possible because of their non-indexed, non-optimized, and non-gamified environments”. Ribbonfarm’s Venkatesh Rao has another term for it: “the Cozyweb works on the (human) protocol of everybody cutting-and-pasting bits of text, images, URLs, and screenshots across live streams”. Where Dark Forests are characterised by their intentional withdrawal from social media, the Cozyweb is non-indexable because of a lacking interconnection between material, creating unintentionally disconnected islands populated by isolated communities.

In spite of their differences in intentionality, both terms describe communities that withdraw themselves from participation in the power laws of social platforms. According to New Models’ Carly Busta, and Joshua Citarella, it is through this withdrawal that these new subcultures find the space to develop shared interests and build their collective lore.

What we believe to be most promising about emerging spaces like Trust is their potential to grow into collectively-owned social and cultural institutions built on decentralised infrastructure: democratically governed manifestations of collective interest ranging from political aims to fandoms, contributed to and run by their members.


The characterisation of these communities as isolated islands or dark forests, while useful in describing our current media landscape, sits uneasily with both our experience of participating in Trust, and our hopes for these spaces developing into democratic social institutions which have a relationship and responsibility to the public. As a recent post by Laura Lotti, Sam Hart and Toby Shorin points out, we need new articulations of how decentralised online communities interact with the public, and attached conceptions of the public good, one that as they put forward, should to be framed around positive externalities rather than internal interests. Therefore we believe there is a risk for communities embracing isolating metaphors such as the Dark Forest. Such an embrace could lead to these communities developing similarly to the historical precedent of merchant guilds: rent-seeking online tribes that optimize for internal value creation but with little to no concept of responsibility to the outside or a similar notion of other publics.

As we collectively figure out how to facilitate and evaluate positive externalities, there needs to be possibility for involvement, participation and exchange with other groups and potential future stakeholders; a public outside the dark forest.

A new route to the outside

If we do not involve a public, we risk institutionalising an understanding of collectivity that only includes and reinforces one’s own tribe while furthering collective (often financial) interests - all happening under the false belief that echo chambers are somehow a valid form of political or social engagement. We will now look at the media currently used to facilitate exchange between these private communities and external publics. To what extent are these dominant media models capable of producing positive externalities? And, to what extent might their current affordances be counter-productive to long-term visions of collective agency and public participation expressed in the desire for decentralised and democratic institutions? Currently, many semi-private communities that rely on crowdfunding to finance their activities also use the public media formats of the creator economy, or Clearnet Stadiums; a vast array of podcasts, private newsletters and video-essays built to highlight individual sensibilities and passions. Creator Stadiums give expression to the thoughts and knowledge of the content-creators that use them to communicate with potential and already existing audiences. This sharing of media formats between dark forest communities and Clearnet Stadiums can be traced to their shared emergence on a certain generation of mid-2010s tech platforms such as Patreon, Discord, Substack, and their reliance on the crowdfunding mechanisms that lie at the core of these platforms. Carly Busta and Lil Internet refer to the relationship between these communities and the tech platforms used for public visibility as the interplay between Dark Forests and Clearnet. It is the moving between Dark Forests and Clearnet which makes sheltered conversation and fermentation possible, whilst making the emergent ideas part of a larger discourse that allows new people to discover, critique, or join the project to contribute and decide on internal and public activities. As New Models point out, clearnet platforms are the next-best thing dark forest communities currently have to facilitate interactions with a public, due to their discoverability and low-barrier participation. That private platforms such as Amazon’s Twitch, Twitter and Facebook now constitute the digital equivalents to public spheres is as much a political failure as one of our collective imagination. As the two trajectories of stadiums and their potentially decentralised counterparts progress, we predict that the media and tooling built for the first will be more evidently incompatible with the goals of the second. Certain types of media are ideal for stadiums that have influencers at the heart of their activity, and some media forms are better suited to decentralised institutions with many contributors: a megaphone is not a great tool for having everyone’s voice heard, podcasts are not interactive formats, and video-essays are (currently) not produced together with the audience.

Map inspired by the Eghbal model of open-source communities, “Working in Public: The Making and Maintenance of Open Source Software”, Nadia Eghbal (2020).
Map inspired by the Eghbal model of open-source communities, “Working in Public: The Making and Maintenance of Open Source Software”, Nadia Eghbal (2020).

In result, there is an emerging need to supplement new distributed funding and decentralised governance mechanisms with new routes between private communities and public participation. We need new participatory media formats, as well as changes to existing formats, to better represent and facilitate collective agency and value distribution so that communities can avoid exactly the effects they are trying to escape from; Web 2.0 influencer dynamics and their power laws, where all the value and control flows to a few central nodes. We think of these private platforms as arid wastelands that still need to be ventured into for loot, the rescue of new members and the building of our new vehicles, until we can replace them with decentralised alternatives.

Moving Castles

In order to be a viable alternative to clearnet stadiums, Moving Castles must reflect the following principles. The principles are adapted from the design goals described in Modular Politics.

  1. Collective: Many contributors share control through transparent and real-time governance mechanisms.
  2. Portable: To avoid lock-in mechanisms, Moving Castles have the ability to move freely between platforms, standards and protocols, from private to public, without losing any value, knowledge, or lore in the process.
  3. Modular: Communities should have the ability to construct Moving Castles by creating, importing, and arranging composable parts (such as avatars, props and environments) together as a coherent whole while making these parts available for others to reuse and adapt.
  4. Interoperable: They have the ability to interact with other communities; communicating, playing games, and sharing knowledge & skills in order to help these communities become Moving Castles themselves.
Sketch of a Moving Castle made up of modules, with additional character design by 
Sketch of a Moving Castle made up of modules, with additional character design by Parrr Geng.

Each moving castle can be thought of as a miniverse inhabited and controlled by a community through a number of inputs. Compared to the generalized idea of the metaverse in which one persistent and shared virtual reality allows for much less customisation on a community level, a miniverse is flexible, adaptable and agile. In this way, a Moving Castle’s structure can be thought of as a mix of a Discord server and a Massive Multiplayer Online game (MMO). And, similarly to a Discord server, a Moving Castle has a main space subdivided into rooms or channels, each composed of a theme, lores, rules and potential mechanisms.

Through magic the door is connected to doorways in other locations, effectively making a door with one interior and several possible exteriors, allowing the wizard(s) to travel anywhere they establish a connection and to easily return home while traveling.


In Wynne-Jones’ book (and in Miyazaki’s movie) the doors of the Moving Castle are portals that lead to a variety of changing worlds and contexts. In our miniverse, the doors enable a variety of changing inputs and outputs, from APIs to communication with other Moving Castles. For example, doors that enable web3 integration would allow Moving Castles to become a representative media of DAOs and Multisigs, controlled by members with the inclusion and visualisation of financial mechanisms and the ability to read and write directly from and to permissionless ledgers. By integrating subscriptions or token access for roles, it becomes possible to circumvent services like Patreon and Twitch subscriptions to source funds, while also allowing far greater portability to future decentralised alternatives (that are already emerging). Other doors could lead to friendly communities, allowing free movement of users, items and lore between the castles.

Moving Castles allow the interchangeable combination of parts in a modular fashion, such as for example: A) raw API inputs B) collective governance such as voting C) resulting web3 outputs.
Moving Castles allow the interchangeable combination of parts in a modular fashion, such as for example: A) raw API inputs B) collective governance such as voting C) resulting web3 outputs.

Full customisability is achieved by using real-time rendering softwares, such as Unreal Engine. These allow the community to commission, create or reshape the components of their Moving Castle themselves. As a result, Moving Castles are not hermetically sealed, self-ossifying spaces, but instead can be swapped, combined and merged in a modular fashion. This scrappy, constantly re-assembling amalgamation of rooms, mechanisms, community members and portal doors is what we define as a Moving Castle. As Moving Castles progress in development, a fifth principle might be added to our list of defining criteria: persistence. Persistent miniverses would allow Moving castles to become alive through continuous gameplay, evolving world-states, unstructured social interactions, lore libraries and idle-games for community members at any time. Moving Castles float in the intertidal zone between public and private waters while providing tools for portable self organisation. By adding more functionalities to Moving Castles, and combining them using our modular structures, we hope to see a Cambrian explosion of new egregores — monsters, biospheres, mega-cities, flying machines and vehicles — all built for specific goals and governed by their respective communities and stakeholders. This in turn could lay the foundation for a network of interoperable Moving Castles, all communicating, playing games and bumping into each other - steered by the shouts of their respective communities.

We are moving from an era of centralized, bureaucratic value creation firms to an era of decentralized, permissionless value creation networks. As organizational models change, so too will the intangible cultural artifacts created by these new institutional forms. Brands, narratives, memes—we now choose our own headless gods.


Having now sketched out our vision for what Moving Castles could be, we now turn to the question of how to bring them into existence.

Kitbashed castles in the sky

Over the past year, we produced a series of live streamed media experiments to test the framework of Moving Castles and develop foundational layers of their infrastructure. Developed using Unreal Engine and hosted on Twitch, our Mascot Streams explored new types of collective agency and encouraged community participation using game design and interactive mechanics.

Mascot Stream 3D game world, motion captured mascot, user avatars, user-inputs in game mechanics, and the resulting chaos that ensued. You can find recordings of all previous streams 
Mascot Stream 3D game world, motion captured mascot, user avatars, user-inputs in game mechanics, and the resulting chaos that ensued. You can find recordings of all previous streams here.

Our Mascot Streams mixed elements of game shows, reading groups and networked knowledge production, all facilitated in a 3D game world by a motion captured mascot. The audience could interact with and control objects, avatars and environments through chat-inputs in real-time. Each stream explored parallels between economic and game mechanisms, using those same game mechanics to trigger the reading and discussion of research and writing on new and strange economic forms and phenomena. Each stream additionally tackled a different experiment in interaction design, aimed at allowing for collective involvement in play and worldbuilding through Twitch. These included live gambling via chat to in-stream viewer representation. In our streams we tested a series of experimental functions relating to the underlying infrastructure of our prototypical Moving Castle:

  1. Real-time motion capture of Mascot(s), directly controlled by streamer/author (face and body movements, content of event) and indirectly by audience (appearance)
  2. Remote control of UE4 world (mascot, props and environment with imbued mechanisms) by audience via Twitch chat (governance)
  3. Audience embodied presence in world through 3D avatars controlled by chat-inputs
  4. Persistent tokens used to unlock characters, interactions, and alternative storylines

In all of these experiments, it was necessary to embrace a strategy of kitbashing, a practice in model building and 3D graphics whereby a new model is created by taking pieces out of commercial kits. In lieu of a fully customisable metaverse; a collective virtual shared space where communities can interact virtually while controlling the economic and programmatic logics at the core of their interactions, we have bootstrapped our Moving Castles by combining existing participatory elements and reach from Twitch live-streaming with the modular customisability of game engines.


In doing so, our streams drew inspiration from a broader current of social experiments such as Twitch Plays Pokémon, created by an anonymous Australian programmer in 2014, in which Twitch-users control a character writing game-inputs directly into the chat collectively puppeteering the gameplay through a very crude form of direct democracy known as anarchy mode. Twitch Plays Pokémon demonstrates how game design that facilitates collective coordination can create alternatives to centralised and authored narratives. We took this as a starting point to think about more complex forms of control, input and output to lay the foundation for a media type that can combine collective agency with public participation, all while using (and subverting) the tools offered by an existing platform.


Looking a little further back in media history, formats such as the Danish ‘live one-player multi-platform interactive game show’ HUGO, first broadcast in 1990, managed to successfully exploit a series of existing infrastructures. Using video game consoles, national TV and landline phones to create a new hybrid format, an audience of young gamers could call in via land-line and control a live-streamed game on their TV using their phones as remote gamepads enabling the interaction with the not-yet manifested future of online games. In the absence of an infrastructure that can function as the foundation for the construction of Moving Castles, and taking inspiration from Twitch Plays Pokémon and HUGO, we propose to kitbash new miniverses using a similar form of hybrid media that combines the participation of Twitch live-streaming with game engines. The resulting miniverses are small and independent worlds inhabited and used by communities which already today accept the tradeoffs of simple- and high-latency inputs in favour of customisable, fun, and participatory media. We believe these streams can point at how decentralised narrative production can bridge false divisions between content-consumers and creators by rethinking how media can be produced collectively. Although authored content still plays an important role in these formats, it functions more like raw material for the audience to manipulate. We found that by giving agency to the audience we made it possible for them to take the event into uncharted territories. This brings forward interesting questions of how to balance authorship and community interest, a question we want to explore further in future streams by building more complex mechanics to test different types of collective agency.

Who is steering?

Heroes are powerful. Before you know it, the men and women in the wild-oat patch and their kids and the skills of the makers and the thoughts of the thoughtful and the songs of the singers are all part of it, have all been pressed into service in the tale of the Hero. But it isn’t their story. It’s his.

– The Carrier Bag Theory of Fiction, URSULA K. LE GUIN

Whereas Joseph Campbell’s mythological concept of the Hero’s Journey was the trademark of an era of authored and linear media production, Moving Castles and similar participatory stream formats that are governed and built by thousands of players/agents could become emblematic of a rhizomatic and decentralised approach to narrative production native to real-time rendered media, that is, (paraphrasing Umberto Eco’s words on the Opera Aperta) a type of media that rather than presenting itself in a well-defined and closed manner to its audience, pre-scripted by its composer: “reject the definitive, concluded message and multiply the formal possibilities of the distribution of their elements”. They offer themselves not as finite works which prescribe specific repetition along given structural coordinates but as “open” works, which are brought to their conclusion by the audience at the same time as she experiences them on an aesthetic plane. As Twitch Plays Pokémon grew, the need for more complex governance mechanisms to facilitate collective coordination were implemented and anarchy mode was supplemented with democracy mode. Democracy mode, as the name suggests, averaged inputs rather than following them directly, creating a more coordinated (and some suggest, boring) end result. These learnings, together with other streams such as Built a Trading Bot to let Subscribers Trade $25,000 on Twitch which expand collective agency to also affect external processes, further point towards how an extension of these narrative formats through more complex governance mechanisms - such as those developed by Trust’s own Black Swan and described in Modular Politics - can turn simple user inputs into sophisticated outputs that interact with a number of external APIs and web3 protocols. This combination of decentralised and real time authorship that can influence external processes enables Moving Castles’ collectively authored narratives to also function as community dashboards for collective governance and stakeholder events. Time-specific formats for engaging low-barrier governance over digital commons and decentralised communities.


Moving Castles are collective machines controlled by governance mechanisms and allow low-barrier participation. They are modular so that they can be customized by communities and parts can be shared between them. By kitbashing Moving Castles from already existing game engines and live-streaming platforms we can build them now while ensuring portability, and through pursuing interoperability point towards a federation of Moving Castles linked together. A network of Moving Castles allows us to escape binary thinking of dark forests and clearnets to instead think about many publics as one network. Many organisational miniverses coming together inside one or many interoperable multiverse(s); larger constellations of Moving Castles. Here, communities themselves define the social and technical rules at the core of their institutions while coming together through a federated model to exchange, trade, disagree and collaborate. Our experiments thus far, have barely scratched the surface of the new types of narrative works that combine low-barrier public participation with collective agency. We hope to see similar experiments by other communities moving towards decentralised institutions that are financed, owned and governed by their own members. With Moving Castles we aim to have presented both an organisational metaphor and one of several very needed media-formats to support this development.

Authors note: The piece was edited to add a clarification around the risk of embracing the metaphor of dark forest communities on 13.08.2021

Thank you to the participants of the collective reading group who through their input and ideas vastly improved this piece. A special thanks to @thejaymo @duster @Son La @bunkerheadz1998 @muein @calsbot @h̷a̷y̷w̷i̷r̷e̷z̶ @arp d @Joanna @ráchel @parrr @nickh @wassim. We also want to thank Other Internet Peer Review for their important insights and feedback which helped shape the piece to what it is today. A special thanks to Jay Springett, Kara Kittel, Toby Shorin, Laura Lotti, Sam Hart, Bryan Lehrer, John Palmer, Callil Capuozzo & Kei Kreutler.

Follow the development on our personal accounts: GVN908ARB. If you have feedback or ideas, please get in touch. Join the Discord.