At 3:45 in the afternoon on 9 September 1947, operators on the Harvard Mark II found a moth in Relay #70, Panel F. They pulled it out, taped it into that day's logbook, and wrote beside it: "(moth) in relay… First actual case of bug being found." Grace Hopper retold the story in lectures for decades, and debugging went from a joke to the industry's verb. (moth) in relay, a web work by the London-based artist and geographer Shaheer Tarar, forks one step off that page: suppose the moth had never been taken out. It is still inside the machine, flying autonomously across global internet infrastructure, and each time it locates a node — a data centre, a submarine cable landing station — it browses the live web for information about that facility, hoping to find a way out of the global relay.

The moth's position inside that machine is a textbook specimen of the remainder. The Mark II was a fully formalised device: every relay closure was written into the design, and the machine simply was its enumerated set of possibilities. The moth is not in the enumeration. It is not part of the design and it is not the opposite of the design; it is the bit that formalisation neither did nor could digest. And the engineers' response was the most standard cut in the whole chisel-construct cycle: extract, tape, caption, file. One thing is usually got backwards here — "bug" as engineering slang for a fault long predates 1947; Edison was using it that way in 1878, and the logbook joke depends entirely on the word already existing. So what happened that afternoon was not the coining of a term. It was the fitting of a specimen to a term that was already there. Once the naming closes, the remainder stops being one: from then on every anomaly has a word that owns it, a procedure that owns it, an hour of billable labour that owns it. That page, moth still attached, now sits in the collection of the Smithsonian's National Museum of American History. It has been pinned twice.

What Tarar refuses is the extraction. Leaving the moth inside returns it from the already-construct back into circulation — it becomes, again, the thing with no name, no owning procedure, and no fixed position, still moving. But the sharper move is the terrain it now moves through. The internet presents itself as pure logic: protocols, addresses, routes, a graph with no places in it. Its remainder is precisely its geography — halls, cooling water, expropriated land, the stretch of beach where the cable comes ashore. Tarar's day job is a PhD on internet geographies, alongside teaching at Central Saint Martins, so his moth does not fly a topology; it flies real estate. And at each stop it reads no text the artist wrote. It reads whatever the live web happens to be saying about that facility right now. The work has no fixed content: it is assembled on the spot out of the sediment the infrastructure has left about its own body. Phototaxis is a moth's instinct. A data centre is a building that is lit, climate-held, and sealed against insects. It keeps flying toward a light it can neither enter nor leave.

The exit is never found. That is not a shortfall, it is the structure — a remainder that finds the exit converts instantly into a construct. The moth stays alive because it keeps failing. And there is a loop here that only closed recently: the sentences it reads about data centres are increasingly written by systems housed in data centres. It is looking for a door in the wall of its own stomach. Ghost in the Loop, the commission the piece belongs to, frames its works as no longer made specifically for human audiences — which is exactly this: the network has begun talking to itself at scale, and Tarar has put an illiterate insect inside the monologue and made it read every word.

The already-construct has to be separated out from what is still alive. Photographing data centres was chiselled through years ago: Timo Arnall's Internet Machine (2014) shot inside a Telefónica facility, Ingrid Burrington's Networks of New York (2016) turned the manhole markings of a city into a field guide, Trevor Paglen dove to the tapped undersea cables and pressed the shutter. That line has a name — the infrastructural sublime — and the name closed over it a decade ago. Media archaeology has a name (Zielinski, Parikka). Glitch has a name (Menkman's Glitch Studies Manifesto, 2010). This moth is none of them. It is not a glitch: nothing is broken, the machinery runs exactly as designed. It is not infrastructure photography: nobody goes to the site, the images are surrendered by the network itself. It is not quite media archaeology either, because it does not dig up something that happened — it extrapolates something that did not happen, and keeps that non-event running in the present tense. There is no word for the intersection yet. When one arrives, the piece becomes an example of it.

Seeing it now matters more than seeing it later, and two clocks are running. The first is the naming clock: the work currently hangs in two places at once — inside a commission from The Photographers' Gallery and Goethe-Institut London, and on the 2026 Tiny Awards Site of the Year shortlist, where public voting is open until 25 September. The chisel is raised. It has not come down. The second clock is harder. The piece reads the live web. The online exhibition runs to the end of 2028, but the work in 2028 will not be the work today, because the sentences about these facilities will have turned over by then. Data centres are among the fastest-changing built environments on earth right now; half the nodes the moth visits today will have been expanded, renamed or sold. What you see is this network's description of its own body at this particular moment — and that state is archived nowhere. It is the rare piece whose remainder is eaten by the act of getting round to it later.

Han Qin (秦汉) · @hqinjarsy · SAE Judgment & Aesthetics