…In January 1992, the newly unified German government made almost the entire archive of Stasi reports available to the public [via the Stasi Records Agency]: a hundred and 11 kilometres of files, divided into some 9,000 index headings, covering half a century of surveillance. It was the most radical release of state secrets in history: WikiLeaks on a vast scale.
The Stasi files offer an astonishingly granular picture of life in a dictatorship—how ordinary people act under suspicious eyes. Nearly 300,000 East Germans were working for the Stasi by the time the Berlin Wall fell, in 1989, including some 200,000 inoffizielle Mitarbeiter, or unofficial collaborators, like Salomea Genin. In a population of 16 million, that was one spy for every 50–60 people. In the years since the files were made public, their revelations have derailed political campaigns, tarnished artistic legacies, and exonerated countless citizens who were wrongly accused or imprisoned. Yet some of the files that the Stasi most wanted to hide were never released. In the weeks before the Wall fell, agents destroyed as many documents as they could. Many were pulped, shredded, or burned, and lost forever. But 40–55 million pages were just torn up, and later stuffed in paper sacks.
…Creating the files took hundreds of thousands of spies and informants, but reconstructing them has been left to only a dozen or so archival workers—jigsaw puzzlers of a sort. In the decades since the Wall fell, they’ve reassembled <5% of the torn pages. At this pace, finishing the job will take more than 600 years.
Last fall, the Stasi archive launched a new effort to automate the project, in the hope that the latest scanners and artificial-intelligence programs could accelerate the process.
…Puzzlers are a peculiar breed. They care more about pattern than content, composition than meaning. The shapes they arrange could be pieces of a tattered Rembrandt or a lost Gospel, but the whole matters less than the connection of its parts. Tietze is 65 and has been working in the archive for half his life. Short and round, with thick fingers and a bald head stubbled with gray, he moves with a stiff-jointed deliberation, never taking his eyes off the pieces. He transferred to this job 3 and a half years ago, for health reasons—most archival work requires too much filing and walking around—and has found that it suits him. He has a patient mind and an eye for shape and line. “The room may look chaotic, but developing a theme takes a while”, he said. “You think the corner is missing, and then you see, Oh, it’s there! It’s an ‘Aha!’ experience.”
The scraps on the table had been pulled from a brown paper sack the size of a large trash can. They were of varying colors, weaves, and thicknesses; some were printed on one side, others on both. Stasi agents probably tried to destroy files that were especially incriminating, but they didn’t have time to be too selective; they often just cleared the pages off their desks. Some documents were shredded, but the machines jammed one by one—they weren’t meant for mass destruction. Other documents were ripped into small pieces in order to be pulped, but that took too long. Eventually, the agents just tore pages in half or in quarters and threw them into whatever containers they could find, sometimes mixed with candy wrappers, apple cores, and other garbage. It was exhausting. The agents’ hands cramped and fingers swelled and skin got covered in paper cuts, and, in their haste, they left an inadvertent record of their work. Each sack was like a miniature archaeological site: the scraps were layered inside like potsherds. If Tietze lifted them out in careful handfuls, a few strata at a time, the adjacent pieces often fit together.
Tietze pulled two scraps off the table and laid them alongside each other. Their torn edges matched, but not the typed words along the tear. He shook his head and tried another pair. Same problem.
…In the years since, the reconstructed files have helped trace an alternate history of Germany. They span all 4 decades of the GDR, Hovestädt says, and cover everything from the Stasi’s investigation of a Nazi war criminal to agents’ infiltration of East and West German peace movements. They describe the persecution of prominent dissidents like Robert Havemann and Stefan Heym, and doping practices among East German athletes. They report on the activities of the West German terrorist Silke Maier-Witt, a member of the Baader-Meinhof gang who went into hiding in East Germany, and on an informant known as “Schäfer”, who infiltrated dissident groups in the G.D.R. The extent of Stasi spying came as a shock to Tietze at first, though he had lived in its midst most of his life. Yet he radiates no sense of impassioned purpose. He just comes to the office day after day, like the Stasi before him, and methodically reassembles what they destroyed.
…Tietze joined the torn halves with a thin strip of clear archival tape—the word Mittag came together along the tear—then flipped the page over and taped the other side. Working steadily like this for a year, he could piece together 2–3,000 pages. All told, the puzzlers at the archive have reconstructed more than 1.7 million pages—both an astonishing feat and an undeniable failure. More than 15,000 sacks of torn files remain. In 1995, when the project was launched, it had a team of about 50 puzzlers. By 2006, the number had dwindled to a handful, as members retired or were reassigned to other agencies. It was clear, by then, that reconstructing files by hand was a fool’s errand. What was needed was a puzzling machine.
…“There were reports on television about a small team manually reconstructing the files”, Nickolay told me. “So I thought, This is a very interesting field for machine vision.” At the time, Nickolay was a lead engineer at a member institute of the Fraunhofer-Gesellschaft, the German technology giant that helped invent the MP3. With the right scanner and software, he reckoned, a computer could identify the fragments of a page and piece them together digitally. The human puzzlers at the archive could work only with documents torn into fewer than 8 parts. They lifted out the biggest scraps and left the small ones behind—often more than half the contents of a sack. A computer could do better, Nickolay believed. It could reconstruct pages from even the smallest fragments, and search for images of missing pieces from other sacks. You just had to scan the fragments and save the images in a database.
The reality proved more frustrating. It took 5 years for the Stasi archive just to respond to Nickolay’s proposal. By 2003, the Fraunhofer team had performed a feasibility study and created a prototype program, later dubbed the e-Puzzler, that could reconstruct pages torn into as many as 10 pieces. But it was another 3 years before the project was funded—a delay that Nickolay blames on a change in government. Then the team’s industrial partner, a subsidiary of Lufthansa, which had been tasked with designing the scanner for the project, dropped out. Scanning was supposed to be the easy part—even some home offices had high-resolution scanners by then. But the pieces had to be scanned on both sides simultaneously, with extreme precision. For images to fit together, their color and texture had to match perfectly, their edges align to within a pixel’s width. “Normal scanners can’t do that”, Nickolay said. “And, when we looked around, we realized that no scanner in the world could.”
The Fraunhofer team eventually found a scanner that could be retrofitted to do the job. But it couldn’t handle large batches of material. By 2014, the team had reassembled only 23 sacks of documents. It was an impressive achievement in its way—the e-Puzzler could now reconstruct pages torn into more than a hundred pieces—but the team had expected to reconstruct 4 hundred sacks. After the project came to a halt, in 2014, Fraunhofer declared it “successfully completed.” Others disagree. As a Stasi archivist put it, “15,000 bags, 23 reconstructed—you can’t call that a success.”
…With a single scanner and a team of 8 workers, the archive in Cologne has pieced together tens of thousands of fragments in the past two and a half years. Yet the scanner and software were never really the problem at the Stasi archive. The original e-Puzzler was already better than people at reconstructing files. It just wasn’t much faster. The fragments still had to be lifted from a sack, picked apart, unfolded, and flattened on the glass to be properly scanned. If the average worker needs 5 minutes to place and scan 50 fragments, scanning every scrap in the Stasi archive will take close to a million hours.
…A similar method [as the Herculaneum papyri] could theoretically be used to digitally unfold the Stasi fragments. But for now the work still has to be done manually. No mechanical press or roller, no clever prosthesis can do it with the necessary accuracy. “We need a robot hand that doesn’t exist”, Schneider said.
As long as there are torn files left in sacks, Hovestädt says, the Stasi archive will piece them back together. In September 2023, the archive put out a call for proposals to relaunch the digital reconstruction project. MusterFabrik was among the companies that were subsequently invited to present a proposal in person. The winner has yet to be chosen.
[Do they need a better robot hand… or better robot software? What about a shotgun photography tunnel system like the one Vernor Vinge proposed in Rainbows End?]