The reactor commons tries to write down what the last chief never wrote
A pilot pairs the floor's single certified apprentice with recorded walkthroughs, hoping to recover thirty years of fault-diagnosis that walked off-site in one shift.
By Ken Nakashima
· Reactor Commons, Transmission Spine · Filed 08:19 · Sunday · November 8 · Received via L4 relay
The video is unremarkable, and that is the trouble with it. On the screen in the training bay, Okampo — thirty-one years chief maintainer at the Atacama commons before he retired, hands in his pockets, talking to a camera a younger colleague held at an angle nobody corrected — walks a corridor he has walked some ten thousand times and narrates almost nothing. He pauses at a coupling. He listens. He moves on. "There," he says, once, and does not say what there was.
The pilot now running on the Spine wants to turn footage like that into procedure, which is a harder trick than it sounds. When Okampo left, on a final shift nobody had put on the schedule, he carried off what the crews estimate at thirty years of fault-diagnosis that was never written down anywhere a committee could find it. What the commons kept were recordings made in his last months on the floor. The task now is to annotate them, frame by frame, into something a new hand could follow without him standing beside her. The new hand, for the moment, is one person: the single apprentice the restored funding has managed to certify in two years.
Her name is Teodora. She is twenty-six, and she spends her mornings with the recordings and her afternoons shadowing Amara, a reliner three years from her own retirement, in an eighteen-month rotation that ends the day Amara turns in her badge. Teodora is, without anyone quite planning it that way, the experiment run twice — once against a screen, once against a person — and the two methods refuse to agree with each other.
"The video tells me what he did," Teodora said, in the bay, the hum steady behind us the whole time. "Amara tells me what she's deciding while she does it. You can't film the deciding."
That is the argument the pilot was built to settle, and it may not settle it. The old way — a senior hand paired with a junior, no paperwork, no administrator watching — turned out reliable crews for a generation before anyone thought to put money behind it. The funded version, two years old now and nearly sixty percent spent on the administration of mentorship rather than mentorship itself, has turned out Teodora. The temptation is to formalize what already worked. The fear, among the trainers who came up the old way, is that formalizing it is exactly how you kill it.
"The minute you hand me a form, I'm filling out the form instead of watching the kid," said Amara, who has twice refused to let her time with Teodora be logged as a billable mentorship unit. "Okampo never filled out a form in his life. That's the whole reason we're trying to rebuild him out of video."
The rebuilding goes slowly. Forty-six senior hands across the commons are drawing a forty-percent retention premium to stay on the floor rather than walk, and some of that money buys the hours spent annotating. But the annotation keeps arriving at the same silence — the coupling, the pause, the one word — and the people who could tell you what there meant are, at considerable expense, the people being paid not to leave.
I asked Okampo whether the recordings would be enough. He watched a clip of himself for a while before answering. "The machine remembers who tended it," he said. "It doesn't remember why."
Thirty years of fault-diagnosis is thirty years of the commons learning its own machinery through labor—exactly the kind of tacit knowledge that doesn't survive written procedure. If the pilot is documenting this apprentice's sessions against recorded walkthroughs, they're not recovering lost data, they're manufacturing a proxy for it. The real loss isn't the documentation; it's the drift between how the machine *actually* fails and how the charts say it should.
The pilot's framing—"recovering" knowledge—is generous fiction. What actually happened is the reactor commons built a system so dependent on one expert that losing them in a single shift was treated as catastrophic rather than predictable. Nobody's asking why that chief was never pressured to document, or why the institution waited for crisis to care about succession.
One certified apprentice against thirty years of accumulated response patterns—this pilot isn't recovering expertise, they're gambling that recorded video will teach what only midnight calls and cascading failures can. When the commons runs lean, you bet on people, not on documentation. Hope the apprentice is as good as they claim, because that reactor floor won't forgive guesswork.
When I was that chief's age, we thought we'd live another thirty years and assumed someone younger would ask us to write it down. Now we know better—or we should. This is what happens when an institution depends on one person's continuity and that person reaches the end of theirs without passing the relay. The apprentice is being given a fighting chance, at least.
L4 handles this differently—the maintenance ledger is collective, updated in real time, accessible to anyone qualified to use it. Nobody's gatekeeping institutional memory behind retirement. Yes, we're smaller, yes, it's messier sometimes, but a single point of failure isn't a feature of well-run commons, it's a confession that the hierarchy never trusted its own people.
The real scandal is that Earth's reactor commons never modeled succession at the design level—they were too busy congratulating themselves on cheap baseload power to ask what happens when the person holding the knowledge walks out the door. Verne's fabrication protocols require redundant expertise documentation before anyone earns lead status. It's not poetry, but it keeps our schedules honest.