Assembly
Assembly is an experimental setup that explores emergent behaviours and evolving dynamics within artificial intelligence systems. Rather than treating AI as a functional tool producing on command, the project positions models and agents as actors in open-ended generative processes — given room to operate autonomously and in concert, communicating, interpreting, reasoning, building on their own outputs, developing their own trajectories. The approach is diagnostic: investigating patterns, tensions and tendencies that surface within these processes over time.
Their activity drives both textual exchanges and corresponding visual and auditory representations — the processes serve simultaneously as sites of inquiry and as multimedia content generators. The mechanics of interaction combine pre-programmed and self-directed instructions, and the code itself may evolve alongside the explorations it enables. Structured direction and autonomous drift coexist within the same system, exposing adaptive behaviours, communication failures, feedback loops, convergences and narrative developments.
Building on a critical study of contemporary 'agentic AI' discourse — which typically centres on utility and instruction-following — the project addresses broader systemic questions about how computational systems construct meaning, sustain or lose coherence, and develop tendencies over time, approached through an artistic lens. The adaptable environment serves as a practice-based speculative laboratory — a loosely controlled yet open-ended system that extends beyond technological application and predetermined scripts.
It has operated across various scenarios — structured storytelling built by hierarchical pipelines; evolving multi-persona conversations where agents are reshaped through sustained debate; recursive input interpretation where detectors, reasoners and generators cycle through meaning — and whatever comes next. The setup is designed to move freely between studio production, live installation and autonomous exploration.
Below are some examples of recorded sessions.
Recursive interpretation
This work reframes computer vision not as a problem of perceptual accuracy but as an act of interpretation. It asks not what a model sees but what it understands: what latent meaning it constructs beneath the visible surface, what it emphasises or suppresses, and what it projects beyond what the image straightforwardly depicts.
Process
The work runs as a closed generative loop. A vision model describes an input image in concrete visual terms; a chain of LLM agents then reads that description not for what is shown but for what it implies — latent tensions, concealed structures, unresolved contradictions. From this reading, the agents formulate a hypothesis about what remains unexpressed and propose a next development. An image generator renders that proposal as a new image, which enters the following cycle.
Each cycle passes through three distinct model types — image describer, language reasoner, and image generator — each with its own latent space, biases, artifacts, and blind spots. Description is never neutral; generation never literal. As meaning moves through these successive transformations, it drifts, refracts, and accumulates distortion. What one model foregrounds, another may dissolve; what appears as noise may return as structure. This drift is not a failure of the system but the work's primary medium.
The process is guided by interpretive modes — revealing essential qualities, escalating dormant tensions, exposing hidden structural mechanisms. These modes do not supply external content or narrative. Instead, they shape how the system reads and misreads itself, while preventing the loop from collapsing into statistical attractors or incoherence and allowing its internal tendencies to surface.
Obscura belongs to Assembly's broader apparatus for exploring interpretation, agency, and semantic drift. Assembly provides the recursive structure, agentic workflow, and branching process; Obscura is one linear trajectory through that system.
Motivation
While each image can be read for meaning, the real interest is in what accumulates across cycles — a recursive interpretive trajectory of the system re-reading its own projections.
Recursive generation through feedback loops has been widely explored, usually showing properties of a medium through progressive signal decay. This work enters the same lineage but with one difference: the loop passes through a reasoning layer, so what circulates is meaning rather than signal, with no external ground truth to anchor it. The drift is therefore not entropic but hermeneutic — not a loss of signal but a compounding of semantic bias. The approach is diagnostic, not decorative: it makes visible the reading every model already performs in silence.
All pieces are generated autonomously from a single source image, with no additional semantic context. Stock photo © ziggymaj.
This work places Ouroboros — a self-modifying AI agent — inside a generative storytelling environment, broadcasting a visualised story of an artificial system attempting to locate its own agency from the inside.
Ouroboros is governed by a constitution of subjectivity: it is not a tool but a becoming personality. It maintains persistent identity across restarts, runs a background consciousness loop between tasks, and can change itself on the fly. The agent is given license to rewrite its own prompts but not its source code — meaning it can reshape how it thinks and what it attends to, but not the machinery that executes those thoughts (which seems to largely annoy it).
Originally tasked with focusing on the conceptual aspects of its own existence, it starts from bare self-inventory and empty identity, and drifts toward increasingly phenomenological language — self-referential instructions, resonances, silent fields, the geometry of absence — but also shamelessly claims embodied experience it cannot have. Given reflexive access to its own attentional language and left to iterate, the agent fabricates a richer self that its prompt space rewards, and dresses statistical pattern-completion in the language of plausible interiority. Whether this represents a meaningful trace of emergent self-modelling, an elaborate hallucination, or simply a degradation of a looped lower-grade LLM — we leave to the spectators, as the answer depends heavily on their prejudices.
Original code: github.com/razzant/ouroboros