• About / CV
  • Projects
    course
    software-system
    speculative-architecture
    talk
    video-essay
    video-game
    • Mountain of Many Voices
      hypertext narrative, installation, and performance, 2025
    • Go To Sea
      point-n-click video game, 2023
    workshop
    writing

    Mountain of Many Voices

    March

    2025

    2025–2026 · Hypertext narrative, installation, and performance

    Mountain of Many Voices, installation view at Times Museum

    Installation view, Times Museum, Guangzhou, March 2025. Photograph: Pengjing (彭婧).

    A hypertext narrative with shared memory

    Mountain of Many Voices (MoMV) takes the form of a hypertext narrative, encountered through installations and performances. Players return to a childhood home altered by urbanization, then ascend a mountain where personal recollections mingle with memories that have no single narrator. Choices made along the way are retained and can return within later passages, shaping the world encountered by other players.

    It is an experiment, or a work of research-creation, on the conceptualization of technical memory, concerning our ability to “return”: an investigation of the conditions in which the past could live. The research and the creation braid into each other, thus producing a myriad of different outcomes, such as installations, performances, essays, software systems, and diagrams.

    The work asks how encounters with a computational system can become conditions of what it does next. In the interactive narrative, its writers establish the story’s initial world and the procedures through which earlier choices return. A large language model is used very sparingly to compose selected records into new passages through interpretive frames authored for the work. As a result, each subsequent player enters a world with an accumulated history and leaves traces that may affect its continuation.

    First commissioned by and presented at Times Museum in 2025, MoMV was subsequently adapted for performances at Hart Haus, Hong Kong, and Asia Art Archive in America, Brooklyn. The English-language version presented through Rhizome’s Counterstructural Commons residency with Mozilla Foundation in May 2026 locates the story among car factories, diners, and arcades in a town marked by industrial decline.

    A black monospaced game screen shows an ASCII drawing of a low house above the heading “One: Return”; the text describes Shane returning home at age thirty in 1996, with green choice fragments below.

    zzyw, Mountain of Many Voices, 2025, Chapter One landing screen from the English-language version, dimensions variable. Courtesy of the artists. The opening screen establishes the narrative world by situating Shane, a thirty-year-old protagonist, as he returns to a childhood home forgotten by urbanization.

    Generated passage in the game interface

    A generated passage appears as moving, wavy text within the authored narrative.

    At the end of a traversal, the participant’s path is combined with artist-designed ASCII patterns to produce a talisman. In the installation, a thermal printer gives this record a temporary material form.

    Two images are shown side by side. The left image is a black screen with a tall colored ASCII talisman; the right image is a museum installation photograph showing a thermal printer producing a talisman print on a pedestal.

    zzyw, Mountain of Many Voices, 2025, generated and printed talisman, dimensions variable. Courtesy of the artists. (a) Screenshot of the generated ASCII talisman from the English-language version; (b) thermal talisman print in the Times Museum installation, Guangzhou, March 2025. Panel (b) image copyright © Pengjing (彭婧).

    Talisman prints

    Talisman prints from the installation.

    How retained choices shape later encounters

    Infrastructure is the ground on which social and cultural life takes place, and its design shapes what can be thought and done within it.

    We could borrow the soil analogy here to think through the temporal dimension. Soil accumulates the effects of what has lived in it (living plants change its composition, dead plants decompose and become nutrients, etc.) and conditions what can grow next. MoMV approaches technical memory through this relation between accumulated history and future possibility. Memory here—or, in a phenomenological register, “retention”—plays a central role as the “ground.”

    Memory as storage and memory as condition

    Memory as storage / memory as condition. Slide 3 of the Rhizome/Mozilla presentation, May 16, 2026; slide area reproduced without speaker notes.

    In MoMV, memory participates in the composition of the world a player encounters. An earlier selection can return as the atmosphere of a scene, an object, or a character’s response. Its significance develops through the later situation in which it appears and through the player’s interpretation. People who never meet can therefore participate in the formation of a shared narrative world.

    We describe this relation as:

    “Memory is identity; identity is working memory.”

    The work acquires a history through the encounters it retains. Its governing procedures remain relatively stable during a deployment, while the memory they draw upon changes. What differentiates one state of the work from another is partly what has happened within it and how those traces enter its next operation.

    Accumulation cone showing successive encounters and retained history

    Accumulation cone: successive encounters build on retained history. At each generation step, authored retrieval selects which records enter the context for composition.

    Since the current paradigm uses the database as the technical form of retention, the project concerns a new paradigm of the database—or something beyond the database that does more than serve data. The questions concern:

    1. What gets stored;
    2. How they are stored (here, the identity of “they” is already being transformed);
    3. How the past (we can no longer use “them” to refer to a stable signified, as they might have changed identity by this point) affects the future.

    This relation depends on how the system carries traces (past inputs) forward. New records are appended to the database, preserving the order of recorded events. Retrieval can follow one participant across a traversal or gather different participants’ responses to a particular question. The writers specify these relations in a custom editor, alongside fixed narrative passages and the points at which generation can occur.

    A browser-based editor shows multiple story blocks, pre-authored options and a dynamic prompt containing placeholder queries for story-so-far and player decisions.

    zzyw, Mountain of Many Voices, 2025, custom backend authoring interface, dimensions variable. Courtesy of the artists.

    Append-only writing and relational reading

    Append-only writing and relational reading: two forms of interacting with the past/database. Extracted from the Rhizome/Mozilla presentation, May 16, 2026.

    An interpretive frame directs how the retrieved material enters the current scene. In the English-language version, a choice made in Chapter Two, “The Yard,” returns in Chapter Three, “Spin-City Records,” through a frame drawing on Jungian psychology. Elsewhere, accumulated choices from different players are composed through an I Ching frame. These frames give the language model a situated compositional role within the work. Its generated text becomes available for a player to read, interpret, and respond to; a partial record of that response can then enter another encounter.

    Two black monospaced narrative screens are shown side by side. The left screen shows Chapter Two, The Yard, with green choice options; the right screen shows Chapter Three, Spin-City Records, with a grey generated passage below green choice text.

    zzyw, Mountain of Many Voices, 2025, paired player-journey screens from the English-language version, dimensions variable. Courtesy of the artists. (a) Chapter Two choice-point at the factory yard; (b) Chapter Three generated “wavy text” passage, partly shaped by the player’s Chapter Two choices.

    From dynamic interaction to history-dependent recursion.

    From dynamic interaction to history-dependent recursion. (a) Cornock and Edmonds’s dynamic-interactive system, redrawn from their Fig. 1(c) [Cornock and Edmonds 1973]: environment (E) and time (T) condition the artwork (A), while participant (P) output returns to A as feedback. (b) MoMV’s extension of this relation: artwork (A) contains retained history (H) and the composition process (C), while participant (P) remains outside A. Within A, relationally selected records from retained history (H) enter the composition process (C), which configures the narrative world presented to P; a partial trace of P’s response returns to H and may condition a subsequent composition. A’s authored procedures remain relatively stable across this relay.

    Conceptual Inspirations

    Current discussions on artificial intelligence and creativity have long been trapped in polarized narratives: one side views AI as a threat that can supplant human subjectivity, while the other extols its instrumental rationality for eliminating the drudgery of creative labor (e.g., OpenAI, 2023). However, neither of these stances addresses the essence of technology as a “mode of revealing” — as Heidegger put it, technology, through Gestell, reduces the world to calculable resources while simultaneously concealing other potential modes of unconcealment (1954/1977). The epistemological crisis of AI is further rooted in colonial logic: from cybernetics erasing the boundaries between organism and machine (Hui, 2019) to reducing meaning to algorithmic “calculation” (B.C. Smith, 2019), its development has always relied on exploitative labor chains, environmental violence, and data colonialism (Crawford, 2021). We reject the binary “tool-dominion” framework and propose to reconstruct our relationship with AI through a dual practice of “Demystification” and “Remystification” — the former exposing the political and material costs hidden within technology’s black box, and the latter embracing incommensurable ambiguity to transform it into a resource for ethical creativity.

    In past projects, we have practiced Demystification through critical technological education, whereas Mountain of Many Voices is a radical experiment in Remystification: when mainstream discourse treats AI outputs as commands or knowledge, we recast them as divinatory, ambiguous revelations that demand human interpretation to confer meaning. Beyond theoretical critique, this choice also stems from our commitment to narrative — whether through video games or interactive ceremonies, we have consistently explored how collective creation can convey a generation’s spiritual map. Players’ inputs become seeds that return to reshape the worlds within the narrative, and the interplay among players evolves into a collective reconfiguration, with large language models generating and inserting symbolic, ambiguous narrative elements behind the scenes. Participants are required to interpret this ambiguity within their own contexts, culminating in a ritual display of collective talismans. The gradual fusion and dissolution of these talismans serves both as a resistance to the deterministic logic of technology and as a tribute to the inexhaustible nature of the human spirit.

    This project is also a continuation of our “Computational Haze” series, an effort to reconstruct a poetic vision of human–machine symbiosis through recursive narratives and ambiguous interactions in an era dominated by automated software. If AI’s “rationality” is fated to collude with colonial modernity—manifested, for instance, in the hegemonic gaze embedded within its training data—then Remystification is a struggle to reclaim the right to interpretation. It does not deny the materiality of technology but insists on sowing a multiplicity of universes within the fissures of its algorithms.

    Research statement: memory and agency within systems

    Mountain of Many Voices (MoMV) is a research-creation artifact.

    The project concerns time, memory, systems, and configuration within cybernetic systems.

    Cybernetics, especially second-order cybernetics, gives us a way to describe the condition in which we live: inside technological systems that we also observe, inhabit, and alter. If we say that software has become infrastructure — if by this we mean that it has become the ground, the soil that conditions what can exist, live, and grow — then we are saying more than that software, and to an extent the net, is where culture and, more fundamentally, thought happen. We are also saying that the ways these structures are designed and engineered — their forms and paradigms — become the limits of how rich, open, and generative thought and culture can be. The question of how anything — freedom, agency, the ability to alter the trajectory of the system we live with and within — can still happen becomes urgent.

    LLMs intensify a familiar software paradigm: the transformation of human intention into system-readable input, and the return of system-generated possibilities as if they were neutral options. We try to approach this question through creation as a way of thinking, with making and coding as a kind of notebook. Through this process, we ask: what kind of design could preserve contingency — that is, the will and intention to alter the trajectories of systems — as produced by users, players, and all those who use and inhabit the system?

    These questions have to do with time: with the temporal dimensions of systems, with what has been accumulated and produced, and with how we deal with them.

    This leads us to think about a new kind of machine, one that treats “memory” differently. Memory is never inert, stored as a mere pile of bits or “data,” a quantitative accumulation that can be seen as discrete, queryable units like numbers.

    Contrary to the popular notion, memory is not inert, standing still in the past as a relic. Memory becomes the ground for what comes next, and it is constantly reshaped by what comes next. Memory is identity; identity is working memory.

    We ask: what if software had a more intimate and functional relationship with its memory? What if memory did not stand outside the operation of the system, but became the organizing procedure of the system?

    MoMV is itself one way of asking that question through creation. It proposes a novel architecture that gives AI a different role in the system: a mediator, with authorial steering, a node inside the web of meaningfulness. Its output becomes meaningful only when it is taken up by writers and players within the evolving world of the work. AI here is neither the author of the work nor a replacement for human making. It functions as a mediator inside a world shaped by writers, players, and accumulated memory.

    Written for zzyw, Mountain of Many Voices, Rhizome presentation materials, 2026. Final print version. Read the poster PDF.

    Publication: Counterstructural Commons

    Printed copies of Counterstructural Commons

    Printed copies of Counterstructural Commons, published by Rhizome, 2026, with support from Mozilla Foundation. The contribution “Mountain of Many Voices,” by Zhenzhen Qi and Yang Wang, appears on pp. 26–32.

    The publication develops a broader system proposal through the image of a library whose books and shelves can be rearranged by use. It extends the inquiry toward changes in a system’s computational organization, beyond the narrative implementation described above.

    Presentation materials and publications

    • Mountain of Many Voices — final presentation. zzyw, Rhizome/Mozilla, May 16, 2026. Slides and speaker notes on memory, narrative, and the authoring system.
    • Mountain of Many Voices — poster. zzyw, 2026. Research statement and annotated diagram of the encounter–memory cycle.
    • “Mountain of Many Voices,” in Counterstructural Commons. Zhenzhen Qi and Yang Wang (ZZYW), Rhizome, 2026, pp. 26–32 (PDF pages 41–47).

    Performances, Workshops & Talks

    10/2025 — Mountain of Many Voices — Live, participatory storytelling workshop and artist talk, Asia Art Archive in America, Brooklyn, NY, USA.

    08/2025 — Mountain of Many Voices: Reconstruct the Relation to AI, participatory performance and talk, HART Haus, Hong Kong.

    Event flyer for Mountain of Many Voices: Reconstruct the Relation to AI at HART Haus

    06/17/2025, 10:00–11:30 AM — “Mountain of Many Voices: Reorient the Relation to Automation,” guest lecture, School of Experimental and Sci-Tech Arts, Central Academy of Fine Arts (CAFA), Beijing, China. Moderated by Professor Baoyang Chen of the Artificial Intelligence and Data Art Laboratory.

    Guest lecture flyer for Mountain of Many Voices: Reorient the Relation to Automation at the Central Academy of Fine Arts, June 17, 2025, 10:00–11:30 AM

    Exhibition History

    • 10/2025 — Electrified Kids, The 8th Intermedia Art Festival, Tank Shanghai + UFO Terminal, Shanghai, China.
    • 5/2025 — THE BLANC gallery, New York, NY, US.
    • 3/2025 – 6/2025 — AI, as Seen at the End of Ownership, Times Museum, Guangzhou, China.

    All textual and visual materials on this website may be cited, referenced, or reproduced for scholarly, educational, or personal non-commercial use with appropriate attribution. Scraping, harvesting, or extraction of content for any purpose requires explicit prior authorization. Unauthorized scraping, particularly for commercial purposes—including the training of machine learning models—is strictly prohibited.

    © Yang Wang, Zhenzhen Qi 2023 | All Rights Reserved





    PROJECT LIST:
    course
    software-system
    speculative-architecture
    talk
    video-essay
    video-game
    • Mountain of Many Voices
      hypertext narrative, installation, and performance, 2025
    • Go To Sea
      point-n-click video game, 2023
    workshop
    writing