Note on Authorship and Method
Part I · Text: Planning, Composition and Revision
This volume was written between early 2025 and spring 2026 as a collaboration between a human author and a large language model (Anthropic, Claude, Opus series). The division of labour, and the principles that governed it, are set out below. They are recorded not as a disclaimer but as a methodological fact: the conditions under which a text is produced are part of what the text is, and withholding them would be inconsistent with the analytical standards the volume has tried to maintain elsewhere.
Structural design
The architecture of the volume — five civic perspectives converging on a synthesis, framed by a preface and a coda — was established by the author in a working brief prepared before any chapter was drafted. The brief specified the central argument (the reconstitution of the city as a platform of value extraction), the narrative persona (a reluctant expert formed by proximity and duration rather than by disciplinary training), the tonal rules (irony through precision; a maximum of two sentences of visible satire at each chapter opening; the self-correction move), and the standing prohibitions (no idealisation of the past; no rants; no policy recommendations except where their absence would constitute evasion).
The five-perspective structure was chosen to prevent the essay from collapsing into a single emotional register. Each chapter examines the same city through a different functional encounter — residence, commuting, leisure, public space, education — so that the cumulative portrait emerges from the convergence of partial views rather than from a single authoritative position. The decision to maintain indirect specificity — the city identifiable but not foregrounded by name in every paragraph — was taken early and held throughout, on the grounds that the processes described are not unique to one municipality and that explicit naming would invite objections of local circumstance at the expense of structural argument.
Drafting and the role of the language model
Each chapter was preceded by a thematic briefing written by the author: a document setting out the analytical framing, the empirical sources to be incorporated, the tonal register, and the specific rhetorical moves the chapter required. The language model received this briefing together with the project brief and all previously approved text, and produced a first draft. The author then revised the draft, sometimes substantially, accepting, modifying or discarding passages on the basis of accuracy, tonal fidelity to the project’s voice, and argumentative coherence.
The model’s contribution was most productive in three areas: sustaining a consistent prose register across chapters of considerable length; generating the museum-caption register for the plate descriptions, which required a specific combination of formality, irony and compression that benefits from iterative refinement; and identifying, within a given thematic briefing, the structural possibilities for the self-correction move — the moment at which the narrator catches himself and adjusts course.
The model’s contribution required the most correction in two areas: a tendency toward rhetorical closure, which conflicted with the volume’s commitment to leaving discomfort unresolved; and occasional over-distribution of irony, which the project rules confine to designated openings and transitional passages rather than allowing it to diffuse through the analytical body of each chapter.
Revision proceeded in tight cycles. The author pasted approved text independently and requested only the specific changed fragment or citation list — not full file rewrites. This kept revision cycles short and avoided unnecessary regeneration of previously settled material. Version control followed a simple convention (chapter_v01.qmd, chapter_v02.qmd, chapter_FINAL.qmd), with substantive changes between versions documented in commented headers within each file.
Source identification and theoretical register
The theoretical framework draws on urban studies scholarship whose relevance to the volume’s argument was assessed by the author: Lefebvre on the right to the city, Zukin on commercial displacement and authenticity, Augé on non-places, Koolhaas on the spatial logic of late capitalism, Sennett on the ethics of building and dwelling, Harvey on capital accumulation and urban space, and Cócola-Gant on tourism gentrification as a specific mechanism of residential displacement. The selection was the author’s; the model assisted with bibliographic verification and with integrating references into prose as single unremarkable sentences rather than literature-review clusters — a constraint specified in the project brief and enforced throughout.
Empirical data was drawn from institutionally auditable sources — Defensor del Pueblo Andaluz, INE, Idealista, the Junta de Andalucía’s tourism registers, and comparable public instruments — on the principle that the volume’s evidential claims should be verifiable by any reader with access to the same institutional archives. Where data heterogeneity prevented defensible cross-regional comparison (as in the traffic sanctions appendix, where non-Andalusian capitals report under incompatible statistical frameworks), the constraint was stated explicitly within the text rather than concealed behind apparent comprehensiveness. This principle — methodological honesty over the appearance of thoroughness — was one of the project’s earliest commitments and one of its most frequently tested.
Discarded approaches
Several structural and tonal options were considered and rejected during the drafting process. An early version of the coda attempted an aphoristic density that, on review, provided the reader with the satisfaction of stylistic closure — precisely the resolution the section was designed to withhold. A draft of the commuter’s chapter experimented with sustained comic exaggeration, which was reduced after it became clear that the darkly comic register worked only when grounded in specific, verifiable observation. The glossary was originally conceived as a more extensive apparatus; a first tranche was completed (with entries for threshold and perimeter), and a second tranche (including infrastructure of affordable sociability) was deferred to align with the chapter in which that concept carries its principal analytical weight. Title variants for the volume and series were generated, assessed and archived in a dedicated file; the final selections prioritised precision of reference over rhetorical attractiveness.
Format and presentation
Two custom SCSS themes — civic-darwinism.scss for the light presentation and civic-darwinism-dark.scss for the dark — are declared in the Quarto project configuration as separate files rather than combined into a single stylesheet. The dark mode was calibrated to preserve the warm, paper-like quality of the light palette rather than to introduce a distinct visual identity: the same fonts, spacing and structural rules apply in both modes, with only the colour values — background, foreground, accent tones (rust, gold) and block backgrounds — recalibrated for legibility on dark ground. Quarto’s native theme toggle, with localStorage persistence, allows the reader to override the system colour-scheme preference.
Part II · Illustrations: Visual Metaphor and Generation
The illustrations in this volume — designated as plates and captioned in a museum-label register — function as a parallel dimension of the argument rather than as decoration or visual relief. Each plate is positioned within or immediately after the passage whose observation it refracts, and its caption operates in a register deliberately distinct from the surrounding prose: formal, terse, and mordantly ironic. The contrast between the caption’s administrative composure and the situation it describes is itself part of the satirical apparatus.
Generation process
Prompting and iteration. Each illustration in this volume was developed iteratively from a written visual brief that specified compositional constraints, narrative function and tonal register, rather than from existing photographic material. Initial prompts translated these briefs into architectural and situational parameters (street typology, degree of crowding, light conditions, palette) and produced a first synthetic image in a documentary–illustrative register. Subsequent versions were generated through successive, narrowly scoped edits, each adjusting a limited set of variables: signage and lettering, pricing conventions and currency, the positioning and posture of figures, or the degree of interior visibility at a threshold. At each stage, the result was evaluated against the brief for legibility (can the scene be read in three seconds?), avoidance of real-world identifiers (no actual brands, buildings or slogans), and consistency with the book’s chromatic scheme.
For the more complex plates, including the real-estate window sequence, the demonstration march and the cover image, this process extended to fine-tuning textual content on placards and posters, ensuring that slogans and micro-texts remained legible, plausible and contextually accurate while avoiding direct reproduction of journalistic sources. Corrections to typographic hierarchy (title, subtitle, author line), language variety and tone were made at the end of the pipeline, once the underlying illustration was fixed, so that text would integrate visually without driving the composition. The resulting images are thus the outcome of practice-led, prompt-based visual research: each plate condenses a series of constrained revisions in which the same underlying scene is re-specified until it satisfies the conceptual, aesthetic and evidential requirements of the project.
Each plate required multiple generation passes. The initial prompt was derived from the scene described or implied in the corresponding chapter text, translated into visual specifications (composition, colour register, level of detail, presence or absence of identifiable features). Candidate images were assessed against three criteria: fidelity to the scene as written, compatibility with the volume’s visual tone (documentary rather than photorealistic; illustrative rather than cartographic), and the capacity of the image to sustain the museum-caption register without the caption having to override or apologise for what the image shows.
Discards. A significant proportion of generated images were discarded. Common reasons for rejection included excessive stylisation (the image read as editorial illustration rather than as documentary observation), insufficient spatial specificity (the scene could have been anywhere), and tonal mismatch (the image was too cheerful, too bleak, or too obviously satirical to support the restrained irony of the caption).
Ethical and representational considerations. No generated image depicts identifiable real individuals. Architectural and urban features are rendered with enough specificity to be recognisable as belonging to a Mediterranean city of the type described, but without reproducing any single identifiable building, street or façade in a manner that would constitute a topographic claim the volume is not in a position to verify. The decision to use generated rather than photographic illustration was taken in part to maintain the indirect specificity that governs the text: the city is this city, but the argument is not confined to it.
Licence and Attribution
© 2026 Miguel Moreno. This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International Licence. You are free to share and adapt the material for non-commercial purposes, provided you give appropriate credit and distribute any derivative work under the same terms. The illustrations, generated as described above, are subject to the same licence as the text they accompany.
The published volume is archived at Zenodo and may be cited using the persistent identifier https://doi.org/10.5281/zenodo.19533278.