Post-NFT digital art. Infrastructure, archiving and recognition II

Persistence. Archiving, custody and the duration of digital art

This research examines, from the perspectives of curatorial practice and collecting, what remains of tokenised digital art after the collapse of the speculative NFT market. The first entry addressed the initial dimension of the triadic framework—provenance, persistence and recognition—through the case of the artist romainiti and the closure of Foundation. That entry concluded by posing two questions: how the durability of a digital work is constructed beyond ad-hoc rescue, and to what extent institutions are equipped to recognise and sustain a heritage that operates under logics distinct from those that have historically organised their work. This second entry in the series addresses the first of those questions.

The conceptual shift regarding the first entry is deliberate. There, persistence appeared reactively: what survives when a platform fails. Here, it is approached proactively: which architectures, practices and institutions produce durability. The distinction is not rhetorical. It alters the vantage point from which the problem is viewed and, consequently, the types of actors that enter the analysis.

This research is authored by Juan A. Esteban. The cases developed—Rhizome ArtBase and the transition from Hic et Nunc to Teia—have been selected for their capacity to illuminate different planes of the same issue: long-term institutional conservation and distributed community custody when no institution assumes the role.

The problem of persistence beyond rescue

When a museum decides to conserve native digital art, it faces a problem absent in other fields: the work it receives may cease to function before the conservator knows how to care for it. Formats change, browsers break compatibility, plug-ins disappear, hardware ages. Faced with this, digital art conservation has developed three strategies that should be distinguished, as each makes different decisions regarding what is considered the work and what is considered its support.

Emulation reconstructs the original technical environment using software that simulates the operating system, browser, or hardware in which the work was conceived. It preserves the experience as it occurred then, at the cost of maintaining a growing fleet of emulated environments that also age. Migration moves the work to contemporary environments—rewriting Flash in HTML5, for example—guaranteeing execution but modifying the work: the question is how much it can be altered without ceasing to be itself. Capture preserves the behaviour of the work as an audiovisual record or as a navigable web archive, replacing live execution with its documentation; it preserves the footprint, not the piece. The three strategies are complementary, and most serious conservation programmes combine them according to the type of work and the intended time horizon.

Rhizome is probably the institution that has developed this problem with the most continuity. ArtBase, its archive of born-digital works, has been operating since 1999 and preserves pieces with a granularity that distinguishes between the original version, archived variants, and accessible copies. That distinction matters because it accepts from the outset something that material conservation rarely makes explicit: that conserving digital art implies managing versions, not objects. The archived work is not the same as that which ran on its original server; it is a documented variant whose genealogy must be registered. Alongside ArtBase, Rhizome has developed web capture tools—Webrecorder first, Browsertrix later—designed specifically for dynamic sites loaded by JavaScript, which traditional web archiving captures poorly or not at all. The instrumental choice is not neutral: it presupposes that a good portion of the digital work of the last two decades lives on the web, and that conserving it requires tools designed for how the web actually functions, not for how it was documented twenty years ago.

Rhizome's trajectory also illustrates a less favourable reality. Conifer, the tool that for years enabled artists, archives and communities to create conservation-grade web captures, has entered a process of progressive replacement in favour of new tools within the Webrecorder ecosystem itself, as part of a transition that the organisation has publicly documented. The operational lesson requires no extrapolation: even preservation infrastructures age and are replaced, meaning that institutional durability is not a stable state but a continuous endeavour that also requires succession planning. Whoever keeps a work alive also keeps the infrastructure that enables its maintenance alive.

Beyond Rhizome, there are two other institutional references that should be developed to understand the landscape. For two decades, Tate has developed a line of Time-Based Media Conservation—video, installation and, subsequently, software-based art—whose work in cataloguing and drafting documentable conservation protocols has produced much of the methodology applied internationally today. The work of Pip Laurenson and her team regarding the conservation of software-based installations, the questions surrounding identity and change in technology-dependent works, and the procedures for documenting behaviour prior to deterioration: this methodological output has come to function as common infrastructure for the field. Furthermore, Tate has undertaken the public dissemination of this methodology through training programmes, international symposia and open-access publications.

ZKM, in Karlsruhe, has maintained a competence centre in media art conservation since the 1990s that combines archiving, technical restoration and theoretical reflection on obsolescence. The institution has worked extensively on complex works with obsolete hardware dependencies, large-format interactive installations and historical software art, which has established it as a continental reference for pieces that no other actor can technically maintain. Its archive and laboratory also function as shared infrastructure for other European institutions, adding a cooperative dimension to conservation that is rarely made explicit.

None of the three institutions—Rhizome, Tate, ZKM—has resolved the problem; rather, they have assumed responsibility for it long enough to accumulate criteria and produce transferable methodology. This accumulation is, in itself, part of what produces durability. The conservation of digital art is not a matter for each institution to resolve in isolation: it is a field of practice constructed collectively, where each successfully resolved case contributes protocol and vocabulary to the rest. Alongside these three references, one must also mention initiatives such as Matters in Media Art—a consortium of Tate, MoMA, SFMOMA, and New Art Trust—and the academic work of figures such as Annet Dekker, whose research into the archiving of net.art and digital art has provided part of the conceptual framework with which the problem is considered today.

Technical Custody and its Economics

When discussing the technical durability of an archive, there are four models that should be distinguished by their economic logic rather than their technical detail. Each responds differently to the fundamental operational question: who pays for this to continue existing one hundred years from now.

The first, and the only one that does not introduce dependencies external to the archive, is fully on-chain. The work—image, generative code, text—lives within the bytecode of the smart contract itself, encoded on the chain. There is no external file to point to: the token and the work are one and the same. As long as the chain exists and continues to produce blocks, the work exists. Projects such as Autoglyphs, Avastars, or a significant portion of Art Blocks Curated operate in this manner, generally under considerable technical constraints—gas costs necessitate lightweight pieces, almost always generative code or minimalist graphics—but offering in exchange the only form of persistence that does not require external actors to keep the archive alive. The economic question does not disappear: it shifts to the maintenance of the chain itself. On robust chains such as Ethereum, that maintenance is guaranteed by the network economy over a very long horizon; on smaller chains, the promise is proportionally more fragile.

The second model is Arweave, which operates under an endowment scheme. The uploader pays once, at the moment of minting, an amount calculated to provide economic incentive to nodes over a very long horizon. The promise is elegant: it does not require renewal, it does not require maintenance, it requires no subsequent action; the system is self-sustaining. The promise has its counterpart. All responsibility is concentrated at the initial moment, which requires trusting an economic projection several decades into the future for a system that has only been in operation for one decade. If the projection was accurate and the network remains active, the promise is fulfilled; if either of the two assumptions fails, there is no subsequent correction mechanism, because the model is built precisely upon the idea that none is needed.

The third model is IPFS. It distributes the file among nodes identified by hash, which protects against the failure of a single server, but requires that someone keep the file active—pinned—on the network. If no one does so, the content may disappear even if the hash remains valid. The economic question shifts: it is no longer who pays once, but who pays continuously. There are institutional pinning services, collectors who assume the cost, and communities that distribute the responsibility. The fragility is not technical but organisational: it depends on there being an actor—individual, institutional, or collective—willing to keep that relationship alive for as long as the work requires.

The fourth model is the artist's own contract, the path least discussed publicly. When the artist deploys their own smart contract, rather than operating within the shared contract of a platform, they retain technical sovereignty over the tokens and metadata. This allows, should the need arise, for the redirection of references, the updating of links, and the management of migrations without depending on the fate of the company hosting the platform. It is the option closest to persistence in the strong sense after the fully on-chain model, but it is also the one that demands the greatest technical and economic capacity from the author: not every artist can assume it, and many should not.

The overarching conclusion is that no single technical custody model resolves the issue of persistence in isolation, with the exception of fully on-chain solutions on sufficiently robust chains, and even those possess practical limitations—volume, work type, and minting costs. Each protects against a specific class of failure while leaving another exposed. True durability is constructed by combining layers: on-chain registration, archiving within a permanent or pinned infrastructure, bespoke contracts where viable, and the explicit acceptance that each of these elements requires an actor to assume responsibility for its maintenance. The economic question admits no general answer: it must be addressed for each work and for every moment of its lifecycle.

Distributed custody and community persistence

Hic et Nunc was, between 2021 and 2022, one of the most interesting NFT platforms in the ecosystem. It operated on Tezos, which allowed artists with limited resources to mint without every technical decision becoming an economic dilemma, and it gathered a community of practice with its own identity, particularly active in generative art, glitch, and small-format experimentation. In November 2021, its founder, Rafael Lima, announced the closure of the hicetnunc.xyz frontend through a brief intervention on social media. The decision did not delete the tokens—the smart contracts continued to exist on Tezos, and the minted works retained their on-chain record—but it did evaporate overnight the interface that allowed them to be discovered, navigated, purchased, and read as works within a curatorial field.

What occurred subsequently is arguably the best-documented community response within the post-NFT ecosystem. Within hours, various community members launched frontend mirrors using the available open-source code. Over the following weeks, several forks consolidated: hicetnunc.art initially, followed by other experimental iterations, and finally teia.art as a stabilised and collectively maintained solution. Teia was established as a community infrastructure with collective governance, open-source code, and an explicit commitment to operational continuity: a budget funded by community contributions, maintenance managed by a rotating core of developers, and operational decisions reached through public deliberation. The tokens that resided on Hic et Nunc remain accessible, navigable, and tradeable via Teia, alongside OBJKT as a broader secondary market within the Tezos ecosystem. The community assumed the function that the original operator had abandoned.

This case clearly illustrates how distributed custody operates. What was preserved in technical terms was that which was already on-chain: contracts, tokens, and referenced files. What was reconstructed was the layer that was not on-chain: the interface, the discovery system, the algorithmic curation of the homepage, and the project's public documentation. That second layer—which the first entry termed operational provenance—is precisely what allows a work to continue circulating as a work rather than being reduced to an orphaned identifier. That this layer could be reconstructed was made possible by two specific conditions: the original frontend code was accessible, and an organised community existed with the technical capacity and willingness to undertake the task. Had either condition been absent, the outcome would have been different.

This introduces the structural limits of the model. Community custody depends upon the continued availability of volunteer maintainers, which renders it particularly exposed to generational shifts: when active members professionalise in other directions, establish their own projects, or simply become fatigued, the infrastructure may be left without support. It also depends upon the health of the underlying network: should activity on Tezos contract significantly, the community persistence of Teia would face a challenge—not of technical existence, as the smart contracts would continue to operate, but of cultural relevance—that it could not resolve from within. Above all, it depends upon the community maintaining the capacity for coordination and governance over a long temporal horizon, which no community model can guarantee in advance.

There is a bridge here to the first entry that warrants explicit mention. The artist romainiti decided, following the closure of Foundation and the rescue migration to Transient, to situate her future continuity on OBJKT, on Tezos. That decision was not arbitrary: it forms part of a broader movement of artists who have migrated to the Tezos ecosystem in search of economic, energy, and community sustainability. The persistence that the Teia case renders visible is the condition of possibility for that migration. Without a community to sustain the legibility of the Hic et Nunc corpus following the collapse of the frontend, Tezos would not be the space it is today. What appears to be a technical decision by the artist—choosing a network, choosing a platform—is in reality a wager on which actor will assume the hermeneutic custody of their work in the long term.

The case teaches something broader about the framework. Persistence is not a property that works possess or lack: it is a distributed labour shared among actors with asymmetric responsibilities. When an institution assumes the role—Rhizome, Tate, ZKM—that labour is professionalised and gains continuity, yet it becomes concentrated within a limited repertoire. When a community assumes it, it gains coverage but becomes dependent upon the health of the community fabric itself. When no actor assumes it, the work evaporates silently, without anyone having to decide so explicitly. The default fragility of digital art is the absence, not the presence, of a custodian.

Return to the framework. From persistence to recognition

The trajectory of this entry has maintained three articulated theses. First: technical persistence, while necessary, is insufficient; the durability of a digital work requires keeping the institutional and hermeneutic layers alive as well, and a work may survive in one while being lost in another. Second: each custody model—fully on-chain, permanent technical, distributed via pinning, sovereign via proprietary contract, institutional, community-based—resolves a specific type of fragility while leaving another exposed; none operates as a universal solution. Third: durability is not an attribute of the work but a continuous practice by specific actors—museums, archives, technical infrastructures, organised communities, active collectors—whose willingness to sustain the work over time is the real condition for the work to continue existing.

This final thesis opens the question that will occupy entry III. If persistence depends on the continuous work of actors who assume its cost, it is necessary to understand what causes those actors to appear, sustain themselves, and possess the resources—symbolic, economic, professional—required to maintain their task. That question is, at its core, the question of recognition. An institution conserves digital art when it considers it valuable enough to allocate budget, personnel, and a continuity plan. A community sustains an infrastructure when its members find meaning—individual and collective—in maintaining it. A collector assumes active custody of their collection when they understand that this function is part of their responsibility, not an optional addition. In all cases, there is a prior process of recognition that renders the maintenance work plausible and sustainable.

Entry III will examine this process through three complementary planes. Institutional acquisition—museums and public collections—as a formal validation device that distinguishes, though not always with the clarity it intends, between acquiring a work, acquiring its support, and acquiring the certificate of ownership for one or the other. The traditional market, specialised criticism, and contemporary curatorship as mediators of criteria that produce a canon before and, often, instead of institutions. And the forms of distributed recognition that operate outside classical circuits—communities, DAOs, decentralised curatorial platforms—which have been decisive in the effective formation of the field during the post-NFT cycle. The three dimensions of the triadic framework—provenance, persistence, recognition—will eventually converge there.