Wandering
    The Archive

    A new form of interacting with historical image material — inspired by Walter Benjamin's ideas of Erkenntnisblitz and Flânerie, merged with the possibilities of high-dimensional embedding spaces.

    Enter the Archive
    Scroll to explore

    The Quiet Repository

    We often imagine the archive as a neutral space — a quiet repository where history simply waits to be discovered.

    But traditional archives are built on something far more rigid: fixed categories, predetermined boxes, institutional hierarchies that decide in advance how the world can be known.

    This is how wandering gets blocked. When every document must fit into a single classification, walls go up between related materials. The unexpected connection — the spark of discovery — becomes nearly impossible.

    You must already know what you're searching for. But what about the things you didn't know existed?

    Filed Under: Marine Corps Records

    Joe Rosenthal's photograph of the flag raising on Iwo Jima — perhaps the most recognizable image of World War II.

    Yet this photograph was originally filed according to pure military logic. Not by its cultural power, not by what it meant to the world, but by the institutional hierarchy of the armed forces.

    Marine Corps records. Branch of service. Date. Geographic theater of operations.

    To find it, you had to think like the military itself — to navigate bureaucratic structures, to know which division, which campaign, which precise moment in the Pacific Theater.

    The image's fame meant nothing to the filing system. You had to know the code.

    A Different Approach

    In the 20th century, Romana Javitz at the New York Public Library tried something different.

    She organized the Picture Collection not by hierarchy, but by human concepts. Keywords like “Labor,” “Hope,” “Solitude.”

    An attempt to make the archive speak the language of ordinary people.

    Taryn Simon, Folder: Broken Objects

    Taryn Simon, Folder: Broken Objects (detail), from the series The Picture Collection, 2012, framed archival inkjet print, 47 × 62 inches (119.4 × 157.5 cm) © Taryn Simon

    Taryn Simon, Folder: Explosions

    Taryn Simon, Folder: Explosions (detail), from the series The Picture Collection, 2012, framed archival inkjet print, 47 × 62 inches (119.4 × 157.5 cm) © Taryn Simon

    Taryn Simon, Folder: Financial Panics

    Taryn Simon, Folder: Financial Panics (detail), from the series The Picture Collection, 2012, framed archival inkjet print, 47 × 62 inches (119.4 × 157.5 cm) © Taryn Simon

    Taryn Simon, Folder: Waiting Rooms

    Taryn Simon, Folder: Waiting Rooms (detail), from the series The Picture Collection, 2012, framed archival inkjet print, 47 × 62 inches (119.4 × 157.5 cm) © Taryn Simon

    The Archive Produces Power

    But even this democratic vision reveals a deeper problem.

    As Michel Foucault warned us, the archive doesn't just store knowledge — it produces power.

    The archivist choosing the keyword becomes a gatekeeper of truth. The same photograph could be filed as “Protest” or “Riot.” The same moment in history becomes “Uprising” or “Disturbance.”

    Each choice opens one door while locking others. Every act of classification excludes — performs one version of reality while silencing alternatives.

    Michel Foucault

    “Animals are divided into: (a) belonging to the Emperor, (b) embalmed, (c) trained, (d) pigs, (e) sirens, (f) fabulous, (g) stray dogs, (h) included in this classification, (i) trembling like crazy, (j) innumerable, (k) drawn with a very fine camelhair brush, (l) et cetera, (m) just broke the vase, (n) from a distance look like flies.”

    — Jorge Luis Borges, “The Analytical Language of John Wilkins” (1942)
    bicycle
    freedom
    truck
    childhood
    sunlight
    labor

    A Topology of Meaning

    Where traditional archives force everything into rigid boxes, embeddings offer a different structure: a continuous space where images and words exist in relation to each other — not by definition, but by learned association.

    This space is not neutral. It's a representation of human cultural knowledge, shaped by the billions of image-text pairs it learned from.

    Press space to explore

    The Archive

    Exploring 1 historical images

    Loading archive...

    Explore the Embedding Space

    Each quadrant controls 64 dimensions. Distance from center = how many. Vertical position = modification strength.

    Searching for "dog"...
    Loading embedding space...

    This is just one way to explore. The full archive offers many more tools.

    Enter the Archive

    Navigating High-Dimensional Space

    Each image in the archive exists as a point in a 256-dimensional embedding space — a mathematical representation of its visual and semantic content.

    Through Principal Component Analysis (PCA), we compress these 256 dimensions into a navigable 2D space, preserving the relationships between images while making them visually explorable.

    Similar images cluster together, forming neighborhoods of visual meaning. As you wander through this space, you encounter unexpected connections — the Erkenntnisblitz that Benjamin described.

    256
    Embedding dimensions
    2D
    PCA projection
    PC1 →
    PC2 ↑
    Loading...

    About the Project

    Wandering The Archive is an experimental interface developed as part of a thesis exploring serendipitous discovery in visual archives. It investigates how embedding spaces can enable new forms of cultural exploration beyond traditional keyword search.

    The archive contains images from historical collections, each embedded in a 256-dimensional space that captures visual and semantic relationships.

    Connect

    contact@example.com
    GitHub Repository
    Enter ArchiveTake Survey
    © 2026 Wandering The Archive
    Built with embedding technology