We Were Here Briefly
2026 · Generative System / Conceptual Art · Concept & Engineering
Overview
The internet produces residue. Not content — residue. Fragments that accumulate in image search results, forum archives, metadata fields, alt text nobody wrote carefully. We Were Here Briefly collects this material daily, without intention, without curation, without a human deciding what counts.
How it works
An automated pipeline scrapes the web using a seed word list that mixes the mundane with the institutional. The day's fragments are condensed into a single sentence using one of five constrained personas, each with explicit anti-instructions: no metaphors about the ocean, no metaphors about stars or time. The sentence becomes a video. The videos datamosh together — not as an effect, but as a process of erasure and accumulation — into a continuous loop. It accumulates whether or not anyone is present to receive it.
It runs. The archive grows.
Architecture
The datamosh is not a filter. The pipeline reads raw MPEG-2 bytes, locates picture start codes at 0x00 0x00 0x01 0x00, checks picture coding type bits, and rewrites I-frames as P-frames — collapsing the distinctions between discrete moments into a single smeared duration. This is the same operation the work performs conceptually: making the boundaries between days harder to find.
The system never crashes on a missing service. Image scraping falls through Bing → Wikimedia → Flickr. Storage falls through Postgres → log.json. The frontend falls through API → direct URL. Cascading fallback everywhere — not as resilience engineering, but as a condition of the work. The archive continues regardless of what fails. Something always gets through.
Related
