AUTOCULTURE ON AN UNDEAD INTERNET
Y7
Passages of this text have been co-authored with DeepSeek and Perplexity AI.
The intention of this text is to outline a thesis (or tell a story) concerning artificial intelligence, network culture, art, content and meaning-making. We hope to provide a quick, non-linear throughline following the growth and development of network culture; the progressive alienation and artificialisation of its constituent parts, and the impact that this has had on semiotics and meaning-making.
Since the start of the year we’ve been trying to take a zoomed-out view of where our areas of research (broadly; internet culture, emergent technology, trend forecasting) might be heading in the future. Heavily influenced by the recent flurry of crypto agent news stories, we seem to be naturally returning to the themes that drew us to AI and its expanded fields in the first place: namely, a fascination with inhuman, non-anthropocentric forms of intelligence and an excitement at the possibility of being able to participate in iterative, generative, networked models of cultural production. We wonder if crypto agents, audio-visual GenAI and large language models might converge (or be converged) into daisy-chained assemblages with various degrees of financial autonomy. This latter part seems to us a profoundly significant notion—capital and agency are presented to us as synonymous, and the introduction of the former as a facet of a generative system makes its capacity to do, to act, to change (for better or for worse) feel infinitely more tangible.
Speculating on these trajectories, we have begun to spin out the emboldening of AI agency through Dead Internet Theory, which is the idea that the majority of activity and content on the internet is already being created and trafficked by AI. Rather than envisaging the dead internet as an interminably buffering mire of collapsed generative models and flatlined feedback loops, we instead wonder if these conditions might become so complex that they give rise to pockets of undead autoculture—of cultural objects (or waves, scenes) produced through the interaction of multiple artificial actors (or factors) with little to no human intervention or initiation. Might the infocene-as-ecosystem pass a tipping point of auto-reproduction and exaggeration of its own defining characteristics? What if the web gives rise to spaces that are not designed for us? What do we as users then become?
Whilst generative AI is certainly the most notable catalyst for these trajectories of artificialisation, automation and alienation, we believe that this is a far longer-reaching tendency traceable through touch points between meaning-making and technology. What follows is an attempt to sketch out an interplay of networks, scales, sensory overloads, volumes, feedback loops, bleached and satiated semantics, all leading to an epoch of autoculture on an undead internet.
THE OTOLITH GROUP, ANATHEMA, 2011
2009-2012 - Semioblitz
A foundational concept for our understanding of the contemporary interplay of technology and culture comes from Mark Fisher by way of Franco Berardi. The term semioblitz is first found in his review of the Otolith Group’s Anathema (2012), a 37-minute video work examining LCD touch screens and communicative capitalism. Speaking about the attention economy that rose with (and through) these technologies, Fisher stated that:
We are exploited now even and especially in our nonwork: our nervous systems work overtime trying to manage the semioblitz but we receive no remuneration for the value that our attention creates.
Being perpetually plugged into multiple communication platforms [...] results in the condition that Linda Stone calls “continuous partial attention.” Here [...] we find ourselves instead in a state of digital jitter, our fingers twitching over the screen, always preparing to make the next click.
13 years later and the shock of the effects produced by communicative technologies seems to have naturalised into a new new normal of second screens, flatlined attention spans and jittery shitpost vernaculars. We have adapted to this restless phase space by introducing new methods of navigation: rather than reading, we sense and scan at accelerated paces metered by media cycles. Semioblitz can be understood as an abstraction produced by our inability to receive information at the scales and speeds at which it is being created and disseminated.
The nodes on the network increase, the market gets flooded, and information’s value—in its broadest conception—depreciates.
Hyperlinks enact a decentralising shift in the field of semiotics, whereby the margins no longer hold the page: one tab recursively leads to another, ad nauseam.
The meme you sent had already been seen elsewhere by the receiver; the news felt old the moment you read it and there was never enough time to attend to the tabs you kept open.
Semio-capital, in fact, is not about the production of material goods, but about the production of psychic stimulation. The mental environment is saturated by signs that create a sort of continuous excitation, a permanent electrocution, which leads the individual mind as well as the collective mind to a state of collapse.
Franco Berardi, Precarious Rhapsody, 2009
2022 - #corecore: Semioblitz as Post-Ironic Medium and Method
#corecore was a TikTok hashtag that briefly garnered mainstream attention at the beginning of 2022, and which took semioblitzed sensing as its form and content. Users of the hashtag would compose random “fifteen-second montages of surreal memes with intense music [that] didn’t have much of a discernible meaning beyond the pleasurable rush of recognizable audiovisual material”. The ethos of this proto-brainrot formalism was laid out in the tag itself: if we take the suffix ‘-core’ (derived from ‘hardcore’) as a form of categorisation used to map and define trends (whilst also subtly intensifying them), we can arrive at the conclusion that #corecore is in fact a category of concentrated categories. In other words, it is a trend whose defining characteristic is the trending of trends. So, whilst there is no immediate implication of an aesthetic (unlike, say, weirdcore), there is an implication of a metamethodology. The emergence of the hashtag can be interpreted as—at its core—a commentary on the scales and speeds of 21st century media infrastructures, which incapacitate users through decision fatigue (you spend 30 minutes searching for something to watch, then you fall asleep) and which accelerate trend cycles with discrete phases into flatlined everything-now continuums.
#corecore can be framed as an expression of a type of existential sickness brought about by network culture: an absurd realisation encountered when trying to conceive of the incomprehensible abstract totality of communicative systems, which either drives the subject into an insensate delirium as they scratch at the walls of their sensorial and psychic prison, or just makes them morosely aware of their impotence in the face of such a vast yet unlocatable infrastructure.
2022 - Hyperimposition: When Does a Collage Become a Sculpture?
When writing our article for Flash Art on #corecore, we coined the term hyperimposition to help us explain the effects and conditions that videos associated with the tag simultaneously interrogated and mobilised. We define hyperimposition as the process that takes place when increases in scale and mass brings about a perception of unintelligibility in the semiotics of media:
If, in superimposition, the composing units being overlaid still retain their existence as discrete and recognizable, then hyperimposition can be conceived of as a process in which the mass and saturation of composing units disintegrate the existing semiotic or formal distinctions altogether.
This can similarly be conceived of as an extension of juxtaposition, in which new meaning is derived from the comparison and contrast of two or more separate things: what if we keep on adding composing units to a juxtaposition? At what point does it lose its capacity to generate intelligible meaning?
2020 - Art Stacks & Squad Wealth
And what happens to the image-maker when the image (both moving and still) exists in the context of myriad neighbouring images, where AI models can produce infinite novel images near-instantaneously? #corecore’s surfacing in the collective consciousness came at the same time that generative AI tools such as Stable Diffusion and Midjourney were becoming increasingly accessible, fundamentally bringing into question the role of the artist. Roughly two years prior to this, pockets of artists, curators and technologists had been pre-empting how we might reconfigure creative practices in the face of such seismic technological shifts.
In July 2020, the Arts Technologies department at the Serpentine Galleries, London, collaborated with creative strategy studio Rival Strategy to launch the first Future Art Ecosystems strategic briefing, which sought to name and frame the new infrastructures being built around artistic practices. One key concept brought about by the wealth of technologies newly available to practitioners was that of the art stack:
Art stacks are artist-led organisations that progressively bring together in-house functions currently distributed between artists, curators, galleries, museums [and] tech companies. The seed of the art stack strategy lies in the need for [artists] to develop integrated studios around DIY approaches to tech. The art stack builds on this position by locating a revenue stream that gives it autonomy from common funding sources in the art industry.
The democratisation of access to tools (and the distribution of resources to learn how to use those tools) increases artists’ capacity to leverage, to achieve greater quantitative results with smaller inputs of energy. Chatbots ghost-write essays and custom Python scripts; text-to-video models produce increasingly uncanny content, text-to-audio models allow musicians to cook up an albums-worth of genrefucked synthesis in a matter of hours: all of these force multipliers inherently shift the position and function of the artist.
Building on top of aspects of Future Art Ecosystems Vol. 1 six months after its release, Squad Wealth—a text co-authored by Sam Hart, Toby Shorin and Laura Lotti—explored similar emerging cultures of proto-institutional group dynamics driven by social and financial precarity, suggesting that the squad could become increasingly central to social, cultural and economic life. Incubated on group chats and Discord servers, squads can generate value through exchanges and shared experiences, which in turn can be translated into tangible outputs such as memes, art, music, startups or side hustles. Here the squad is conceived of not just as a social group, but also as an emerging economic and cultural force perversely engendering both a shift away from neoliberal individualism and a swing towards the incentivisation of social dynamics. This omni-financialising tendency is a recurring theme when discussing the future of the internet—often referred to as Web3—which is generally considered to entail decentralising infrastructural integrations with cryptocurrency and blockchain technologies. These technological and social shifts in agency and leverage are all indicators of a fundamental change in who or what the internet is for and how it is used, and set the scene for an epoch of autoculture.
2024 - The Medium is The Mass, The Medium is The Model: Protocolism
As well as pooling resources in art stacks and squads, some artistic practices engaged with emergent technology are resituating their practices upstream (or deeper into the tech stack). What this means is that—if we momentarily isolate this in visual media—certain artistic practices are no longer concerned with the individual image, but rather find their medium in tweaking the conditions and mechanisms that give rise to images, such as custom datasets and AI models. A crystallisation of this upstream trajectory was vocalised by New Models podcast in their provocation that the art object of Midjourney is not the myriad images it outputs, but instead the model itself. This relocation of practices relinquishes human intervention in certain front-end aspects of the creative process, instead adopting positions that are adapted to the flows of network culture and the force multipliers of neural media.
This movement upstream has also been termed Protocolism by legal scholar, internet activist and artist Primavera De Filippi, who situates this repositioning of artistic practices in a lineage of generative creation that includes John Cage’s 4'33”, Mat Dryhurst and Holly Herndon’s Holly+, and (we would add) Conway’s Game of Life, all of which conceive of the ‘artwork’ as existing prior to the ‘event’: instructions, constraints, models or conditions which give rise to infinitely varying instantiations whose provenance is still identifiable as originating from one place. Similarly, Protocolism establishes a framework of understanding that correctly identifies the ‘artwork’ associated with NFTs as not being the digital images that the term became shorthand for, but instead the smart contract that the image is associated with, or representative of.
Upstream Protocolist practices take an exponentially semioblitzed (and financialised) mediascape as their starting point, employing methods more readily associated with governance and curation to navigate and create within (and on top of) an automated context of maximal content.
2005 - Google Will Eat Itself: Network Auto-Cannibalisation
An anarchic preempting of upstreamed Protocolist practices can be found in the 2005 hacktivist artwork Google Will Eat Itself, which was created by Alessandro Ludovico and UBERMORGEN. The project constructed financial feedback loops geared towards the goal of turning Google into a publicly-owned entity via the creation of a network of hidden websites displaying Google text advertisements. Bots were then used to automatically click on these ads, generating revenue through Google's AdSense program, which in turn was used to automatically buy Google shares that were distributed through a company called GTTP Ltd. (Google To The People). GWEI’s détournement-in-principle manipulates the propulsive insensate flows of capital underpinning the platforms we operate on and through.
2020 - Flight from Clearnet & The Dark Forest Theory of The Internet
The flows of high-frequency extractive financialisation highlighted by GWEI largely follow the traffic of human users congregating on the Clearnet—a term referring to the publicly accessible parts of the internet indexed by traditional search engines. Clearnet is the visible surface of the web that most people use daily for browsing sites, for shopping and for the use of social media. It is the first category of a three-tiered mapping of the internet popularised by Yancey Strickler and Caroline Busta, called The Dark Forest. At the opposite end of the spectrum to Clearnet is the Darknet, which is the parts of the internet that are intentionally hidden and inaccessible through standard web browsers, instead requiring specialised software, configurations, or authorization to access them. The Darknet is a subset of the Deep Web, which includes all parts of the internet not indexed by search engines.
The secret third category in this schema lies somewhere between the Clearnet and the Darknet, and is what Strickler and Busta term The Dark Forest, a term borrowed from Liu Cixin's science fiction novel The Three-Body Problem, in which it is used as a metaphor to describe a universe where civilisations hide in fear of detection by potentially hostile entities. Dark Forest spaces are online communities where users (or squads) can interact anonymously and freely, yet don’t require the specialised software needed to access the Darknet at large. Private Discord servers; invite-only networks, paid newsletters and encrypted group messaging are all places users retreat into to avoid Clearnet hazards such as surveillance, harassment, misinformation, data theft and algorithmic exploitation. In Dark Forest spaces—often used in tandem with Clearnet presences, not as an alternative to them—users prioritise safety and authenticity over visibility.
2025 & 1991 - Non-Player World, Dark Factories & Paris Syndrome: Touching Grass Will Never Be The Same Again.
The impact and effects of communicative technologies, network culture and AI are not confined to the virtual. Our experiences of online spaces frame and inform our offline ones; from the alienation of mediation and the fatigue of semioblitzed maximalism, to the confusion over the authenticity and provenance of content (as well as individuals) online. Similarly, the collapse of global narratives in the past decade—in tandem with the entrenchment of digital infrastructures brought about by COVID—have led to a world that feels both stranger than fiction and as on-the-nose as the most mid Netflix series. The membrane between IRL and URL increasingly feels weak, and the alienation of online spaces across the Dark Forest schema crosses over into a derealised meatspace.
This distrust and destabilisation leads to theories such as the real-life existence of Non-Player Characters, or NPCs, who are the characters in video games not controlled by a player, and whose actions, mannerisms and movements are predetermined by code or AI. But perhaps a more helpful way to frame touchpoints between online and offline is a speculative notion of Non-Player World. Rather than being a simple reformulation of simulation theory (the idea that the world does not really exist) NPW is instead a theory of a very real world becoming increasingly inhuman, giving rise to new spatial typologies directly shaped by the affordances of AI and automation, such as ‘lights out’ dark factories in China. Similarly, we’ve seen the open outcry of 20th Century financial market pits gently fade into the white noise of GPU fans in climate-controlled industrial warehouses, where high-frequency trading algorithms operate at clinically efficient, inhuman speeds. NPW could also be more diffusely conceived of as a real-life vibe of the virtual: a content-pilled lens rendering spaces and places as mere simulacra, and us as tourists in our own reality.
This latter definition is by no means a 21st Century phenomenon. In the 1980s Japanese psychiatrist Hiroaki Ota identified a rare psychological condition affecting Japanese tourists visiting Paris, who experienced severe disappointment or culture shock when their idealised, mediated, virtual impressions of the city were not matched by the reality. Paris Syndrome, as it became known, was characterised by symptoms such as delusions, hallucinations, anxiety, derealisation and dizziness. There are even unconfirmed reports that the Japanese embassy in Paris offers a 24-hour hotline for affected tourists, with treatment involving rest, counseling or repatriation.
What we are trying to convey here is that you can’t just go off-grid (or to Paris) and hope that network culture and its effects just go away. Things like Shrimp Jesus and $FARTCOIN are very real phenomena, and you should take them—and this text—incredibly seriously. We are seeing the unfolding of a viral process through which the world becomes terminally online in ways that bear no relation to screen time metrics. Being online is undergoing a permanent phase shift from verb to noun: it’s becoming a state of mind, an irreversible contagion tearing through the field of semiotics to darken the doorstep of even the most off-grid prepper. Our increasing reliance on the online to facilitate the offline muddies our perception of network cultures’ insides and outsides, presenting us with a very real problem when we identify those same former insides as being increasingly inhuman spaces.
2021 - Dead Internet Theory
The prominence of GenAI in public consciousness has given rise to the folkloric theory that the majority of the internet's content is generated and circulated by AI agents, bots and algorithms, creating cycles of artificial engagement. Targeted advertisements, echo chambers and For You Page algorithms warp and distend online spaces in response to our digital footprint and previous behaviours. AI-generated simulations of human activity turn comment sections into buffering zombified necropolises, and the syntax of artificially-generated viral posts begin to technopomorphically pollute the composition of human-crafted posts. In Dead Internet Theory, the endgame posited is one of collapse: the outputs from AI models currently producing the environmental storytelling of our online experiences are inadvertently included in the datasets of new models, and an ouroboric flatlining of model collapse is supposedly set in motion.
2025 - Undead Internet Theory & Shrimp Jesus
In this instance, the Internet isn’t so much as “dead” as it is undying. What may be considered a virtual ghost town of machine-on-machine interactions could soon be populated with bots carrying out tasks on behalf of humans with other agents.
Again, it’s not so much that the Internet is dead because bots and humans are primarily interacting with bots online. Instead, it’s entering a zombie-like state, bustling, undying cyberspace, where no one really knows if anything is real, yet things seem to keep chugging along.
Eddie Mitchell, March of the Web3 AI Agents: An Undead Internet Theory, CCN, 2024
But perhaps—if we are to take notes from economics—collapses aren’t necessarily definitive. Instead, we might envisage a theory of an undead internet that is simultaneously inhuman and animate, which begins to demonstrate emergent behaviours and nonlinear dynamics as thresholds of artificial complexity, scale and speed are crossed. This is to say that as an ‘internet-of-agents’ grows it is unlikely that it will scale ‘invariantly’ (i.e. maintaining the same properties, behaviors, or patterns regardless of changes in scale) but instead will begin to behave in unexpected ways that seem counterintuitive in comparison with the behaviors observed on smaller scales. Perhaps then, in its increasing complexity and emergence, the internet-of-agents will become a backroom infrastructure of artificial systems within artificial systems, which users in turn navigate by proxy via AI agents.
Might autoculture on an undead internet be defined as the by-products of bot-to-bot interactions that resemble cultural artefacts as we understand them? Might they be the products of models outputting engagement bait for a hallucinated human audience that is simply not there? Think here of an extrapolation of the conditions that gave rise to Shrimp Jesus: a confluence of data scraping, quantitative aesthetics, GenAI, click farms, engagement hacking and FYP algorithms, which converged into latent space embeddings prompted by an aggregated forecast of our desires. This is autoculture as the surfacing of the semioblitzed undertow of network systems.
2035: A Future Network Archeology
As we move into the future, and as meaning continues to lose its meaning through ever-accelerating loops of virtualisation, derealisation and semioblitzed sensing, and as artificial agents’ financialised capacity to intervene in meatspace becomes emboldened, might we see a collapsing of the already porous distinctions between online and offline?
Finally, is it not logical that an undead internet-of-agents (and the autoculture therein) might not be immediately navigable, indexable nor legible to us? If artificial systems give rise to yet more artificial systems created with limited human oversight or intent, and increasingly composed of inhuman typologies, might we begin to work backwards in time through a network archeology that analyses datasets, reverse-engineers protocols, and traces flows of information across an emergent infocene towards autocultural artefacts of ambiguous origin?
Declan Colquitt and Hannah Cobb, working under the pseudonym Y7, are a post-disciplinary duo specializing in theory and audio-visual work, operating out of Salford.