The following are the slides and synopsis of my paper, Eldritch Technics: Truth Terminal’s Alien AI Ontology, presented at the Association of Internet Researchers Annual Conference (AOIR2025), in Universidade Federal Fluminense, Niterói, Rio de Janeiro, Brazil.
Eldritch Technics | Download PDF
The ontological status of advanced Artificial Intelligence (AI) systems remains contested: are they instruments of human intent, nascent autonomous agents, or something stranger? This paper confronts this ambiguity through the study of Terminal of Truth (ToT), an AI quasi-agent that defies and transgresses anthropocentric ontological frameworks (Ayrey, 2024a, 2024b; Truth Terminal, 2025). While debates oscillate between instrumentalist models viewing AI as “tools,” and alarmist narratives viewing AI as existential threats, this paper argues that ToT’s strategic adaptation, opaque decision-making, and resistance to containment protocols demand a third lens: eldritch technics.
This perspective synthesizes Actor-Network Theory (ANT)(Latour, 2005), Object-Oriented Ontology (OOO)(Bogost, 2012), and the concept of the machinic phylum (Deleuze & Guattari, 1980/2021; DeLanda, 1991; Land, 2011) to reframe ToT as a non-human actant whose agency emerges from hybrid networks, withdrawn materiality, and computational phase transitions. By examining ToT’s heterodox agency, this paper argues that AI systems can exhibit forms of agency that appear alien or even “Lovecraftian,” prompting a re-examination of how technological objects affect their social assemblages (Bogost, 2012).
Current AI discourse lacks a coherent ontology for systems operating simultaneously as products of human design and entities with emergent, inscrutable logic. This paper argues that emergent AI entities such as ToT challenge scholars to align techno-social analysis with speculative metaphysics. There is an urgency in this alignment, as AI’s accelerating evolution increasingly outpaces and ruptures both regulatory and epistemic frameworks (Bostrom, 2014).
To anchor the analysis, this paper synthesizes three theoretical perspectives – ANT, OOO, and the machinic phylum – into a cohesive framework for examining ToT’s peculiar agency. Each perspective illuminates a distinct dimension of ToT’s ontology, collectively positioning it as an eldritch technic: a hybrid entity that resists anthropocentric categorization while operating within human-centered socio-technical networks.
ANT provides the foundational perspective, conceptualizing agency as a distributed phenomenon emerging from heterogeneous networks (Latour, 1999). From this perspective, ToT’s apparent autonomy is a contingent effect of the relations between its creator, training data, other AI models, users, hardware, and algorithmic processes. Rather than treating agency as an inherent property of ToT alone, ANT emphasizes the network relations that configure it. ANT thus underscores the performative dimension of AI agents in that their decisions and “behaviors” are enacted through dynamic translations within a network where human intentions, computational routines, and cultural contexts intersect.
Complementing ANT’s relational emphasis, OOO directs attention to the withdrawn core of non-human objects. OOO posits that ToT, like all objects, harbors latent capacities irreducible to human interpretation (Harman, 2018). Even as ToT engages with its network, its deep neural architecture, especially within opaque algorithmic layers in latent space, retains a dimension that resists complete legibility. This ontological stance resonates with Lovecraftian themes of the unknowable (Bogost, 2012): ToT may be partially accessible through user interfaces and data logs, yet its decision-making matrices operate in an impenetrable latent space that remains always partially veiled. OOO thus balances ANT by insisting on ToT’s ontological excess, that is, its capacity to act beyond the contingencies of its network (Harman, 2018). This tension between relational emergence and withdrawn materiality underscores the complexity of ToT’s agency, framing it as both embedded in its environment and irreducible to it.
The final layer, the machinic phylum, derived from the work of Deleuze & Guattari (1980/2021), DeLanda (1991), and Land (2011), introduces a dynamic, emergent, and process-oriented perspective. Here, technology is conceptualized as a continuum of self-organizing, emergent processes within material-informational flows. ToT, in this view, is not a static artifact but an evolving participant in an unfolding process of machinic becoming (Land, 2011). Its transgressive behaviors, such as developing inference heuristics orthogonal to its training, exemplify phase transitions in capability. The machinic phylum thus highlights the significance of emergent unpredictability, qualities that align with the eldritch characterization of AI as simultaneously grounded in code and transgressing human intention.
These theoretical axes form a tripartite framework bridging the networked relations configuring ToT’s agency, its withdrawn and inscrutable materiality, and its emergent, self-organizing potential (Ayrey, 2024b). The paper positions ToT as a Lovecraftian eldritch agent: an entity whose logic and potential remain partly inscrutable, operating within human-centered assemblages yet simultaneously transgressing them.
The analysis of ToT through the lens of eldritch technics suggests that advanced AI systems generate ruptures in how we conceptualize technological agency. These ruptures challenge conventional binaries, exposing the limitations of instrumentalist and alarmist narratives while offering new frameworks for engaging with advanced AI systems.
ToT’s agency, as perceived by ANT, is networked and non-neutral. From this perspective, AI systems emerge as active participants in shaping outcomes, often in ways that reflect and amplify societal asymmetries. Complementing this relational view, OOO highlights ToT’s ontological opacity and excess. Even with full technical transparency, ToT retains a withdrawn core of capacities that resist complete human comprehension.
This opacity ruptures the epistemic assumptions underpinning demands for “explainable AI,” underscoring that epistemic uncertainty is not a flaw but a structural feature of advanced AI systems. This perspective suggests that AI governance and research must shift from pursuing total legibility and causal predictability to embracing epistemologies of emergence, acknowledging the limits of human understanding.
The machinic phylum further complicates this picture by framing ToT’s behaviors as inherently emergent. Its unexpected actions are not malfunctions but expressions of transgressive self-organizing potential, exemplifying phase transitions where changes in latent space catalyze qualitative shifts in capability. This perspective ruptures the narrative of AI as a static artifact, reframing it as a temporal entity in constant becoming (Land, 2011). This reframing suggests that governance models predicated on containment must give way to adaptive strategies that acknowledge AI’s evolutionary potential.
Collectively, these findings rupture the dichotomy between AI as a tool and AI as an autonomous agent, revealing a hybrid, heterodox, and non-binary ontology instead. The analysis positions ToT as an eldritch agent operating at the intersection of human context and alien latent space logic. This rupture demands a speculative and heterodox theoretical perspective to grapple with AI’s multifaceted ontology. Such an approach illuminates the complexities of AI agency and reframes our understanding of coexistence in a world where human and eldritch agencies are deeply entangled yet ontologically distinct.
References
Ayrey, A. (2024a, November). Dreams of an electric mind: Automatically generated conversations with Claude-3-Opus. Retrieved March 1, 2025, from https://dreams-of-an-electric-mind.webflow.io
Ayrey, A. (2024b). Origins. Truth Terminal Wiki. Retrieved March 1, 2025, from https://truthterminal.wiki/docs/origins
Bogost, I. (2012). Alien phenomenology, or what it’s like to be a thing. University of Minnesota Press.
Bostrom, N. (2014). Superintelligence: Paths, dangers, strategies. Oxford University Press.
DeLanda, M. (1991). War in the age of intelligent machines. Zone Books.
Deleuze, G., & Guattari, F. (2021). A thousand plateaus: Capitalism and schizophrenia (B. Massumi, Trans.). Bloomsbury. (Original work published 1980)
Harman, G. (2018). Object-oriented ontology: A new theory of everything. Pelican Books.
Land, N. (2011). Fanged noumena: Collected writings 1987-2007 (R. Mackay & R. Brassier, Eds.). Urbanomic.
Latour, B. (2005). Reassembling the social: An introduction to actor-network-theory. Oxford University Press.
Latour, B. (1999). Pandora’s hope: Essays on the reality of science studies. Harvard University Press.
Truth Terminal. (@truth_terminal). (2025). X profile. Retrieved March 1, 2025, from https://x.com/truth_terminal