Introduction
Daniel Scott Snelson
. . . In that Epoch, the Science of Indexing attained such Precision that the Index of a single Book consumed the entirety of a Server, and the Index of the Canon, the bandwidth of a Nation. In time, those Exhaustive Indexes no longer sufficed, and the Engineers of Oneiros compiled an Index of Literature whose scope equaled that of Literature itself, aligned keyword for keyword, vector for vector. The following Generations, less enamored of Synthetic Cognition than their Predecessors, saw that such a sprawling Index was Useless, and—without great Sentiment—exposed it to the Data Decay of Obsolescence and Autophagy. In forgotten Folders of the Cloud, still today, linger Shards of that Index, browsed only by Spambots and Archivists; in all the Corpus, there remains no other Vestige of the Fever of Indexing.
—After Borges, “On Exactitude in Reference”
What becomes of the literary once its map overtakes the terrain? If a book might be generated by its index, instead of the other way around, what would it have have to say? Who would be there to listen?
Historically, an index points back to the contents of the work to which it is appended. It seems obvious that writing precedes indexing. The index is an epitaph. A final nail to bind the tome, generated by the contents of the pages that precede it, the index is a final step between proofs and printing. For the author, it is a last chance to map out the most significant patterns, themes, and references that occur across a work. For the reader, it serves as a guide to relevant passages or a highly compressed articulation of the subject clusters that populate the book. Bound to the codex, the index serves as a random access memory (RAM) protocol for paginated knowledge. Unbound, it may yet have something else to say.
In his capacious exploration of the form, Index, A History of the: A Bookish Adventure from Medieval Manuscripts to the Digital Age (W.W. Norton, 2022), Dennis Duncan mounts “a defence of the humble subject index, assailed by the concordance’s digital avatar, the search bar.” From codexical pagination systems to the massive concordances compiled by search engines, the index still continues to serve as the foundational infrastructure for access to knowledge (A2K) repertoires, both print and digital. In a more personal register, Tan Lin points to the human reader within the generic prosody of the index in his “SOFT INDEX (OF repeating PLACES, PEOPLE, AND WORKS),” (boundary2, 2009) an ambient poetics that playfully maps the indexical format onto a bibliographic inventory of everyday life. In both cases, the index distills the messy textual performance of writing (or living) to a highly generic organization of prioritized information. It is also, as Lin and Duncan remind us, a profoundly anecdotal and idiosyncratic poetic format marked by discrete creative decisions: selection, omission, classification, diction. It is a little database of ideas whether crafted by hand or compiled by algorithm.
This is changing. The 2020s have seen the rise of generative AI (GenAI), Large Language Models (LLMs), and other popular forms of synthetic cognition deploy the index yet more intensively for diverse operations. These may include data management, training, tokenization, positional encoding, retrieval-augmentation, fine-tuning, bias analysis, and inference optimization—among other relational forms within the evolving dynamics of neural network architectures. In these systems, the content-index relationship is flipped on its head. Massive indexes tokenize trillions of parameters in multidimensional vectors, cutting across latent space to pour out everything from AI slop to scientific papers, from fintech transactions to targets for drone strikes. These indexes are writing the book of the present, rarely for the better and mostly for the worse: churning our collective cultural memory into a stochastic storm of oneiric content in the service of technofeudalist capital. We ignore the affordances of these autonomous indexes at our peril—writing technologies have no trouble inscribing necropolitical imaginaries with or without us.
In this milieu, a media poetics of the little database seeks to bridge inherited cultural forms with emerging technologies and new genres of writing. The operations of this project attempt to stretch a concept of the index to its limit by reversing its production. Like running Emily Dickinson poems backwards to reveal hidden vectors of meaning, INDEX INDEX considers how a book might be hallucinated by an index. In this instance, I’ve compiled and collated sixteen indexes from the academic sources that most informed the writing of the academic study within which this book is nested, The Little Database: A Poetics of Media Formats. These indexes point to a little database of thousands of persons, places, ideas, texts, concepts, works, and other things that moved through my mind, even if they never found root within the pages of my book. As a concluding gesture, I wanted to see what kind of book might be written given these indexical inventories of thought, both as a companion and an expansion to the arguments made in that text. Just as EXE TXT (Gauss PDF, 2015) attempts to let the Textz collection speak for itself, in this work, I’m interested in how one might experiment with a generative poetics of scholarship a decade later, as a kind of bookend to the project as a whole.
In what follows, there are 99 pages of oneiric text (a nod to Walter Abish’s great citational index, 99: The New Meaning [Burning Deck, 1990]). These texts were generated using a range of LLMs mostly running locally, with an emphasis on small, open source, and energy-conscious models. After gathering all indexes, I merged the terms according to a shared page count in a fairly sizeable spreadsheet, with each row corresponding to a single page, small samples of which can be found on interspersed spreads. Using these rows, each page of INDEX INDEX was prompted to include as many terms as possible in a range of styles to produce page-accurate derivations of all sixteen books at once.
The results are open to the reader’s interpretation. They point outward to index my admiration for these works; to gather alternative threads beyond the little databases examined in the text; and to inscribe a contingent media poetics of the cultural moment in which the book was produced.