Chapter 11 — The Caveats · The Relay

The question, when it finally came, was the one Jun had spent his whole career learning how not to answer, and the room wanted an answer so badly he could feel it pulling on him like weather.

"So," the sub-prefect said, with the patience of a man who had sat through forty minutes of qualifiers and intended to leave with a sentence he could repeat. "In plain terms, Analyst. Is the network under coordinated attack — yes or no."

There were perhaps thirty people in the briefing hall at Cardinal Crossing, which was thirty more than usually cared what a consensus analyst thought, and the difference was the dark. Cardinal was the great feed-hub of the inner worlds, the place the network's chatter pooled before it ran back out — a town of relay-masts and reading-rooms and the low constant hum of a civilization talking to itself — and lately the chatter had only one subject. The thirty people had come to hear the man with the instruments tell them whether to be afraid of what they were already, visibly, afraid of.

"The honest answer," Jun said, "is that the data is consistent with several mechanisms, of which coordinated interference is one, and not, on current evidence, the best-supported one—"

"That's not an answer."

"It's the accurate one. If I give you a cleaner sentence than the evidence supports, I'll have done you a disservice that feels, today, like a service—"

"Gods," somebody muttered, not quietly, and the room shifted, and Jun watched thirty frightened people decide, in real time, that the man who wouldn't tell them what to think was less useful than the men who would. He had drawn the curve of that turning. He could not get anyone to sit still for it.

The sub-prefect closed his folder. "When you're sure," he said, making the word an accusation, "the Crossing would be glad to hear it." And the briefing was over, and the thirty went back out into the humming town to get their certainty somewhere it was being given away free.

Jun's field was, in the unkind summary of people who didn't have one, measuring what everyone already knew. He measured belief — which claims spread, how fast, through which mouths, shedding which qualifiers on the way. Twenty years at it, and the charts were beautiful.

The single sentence he could never say in a briefing hall was simpler than the charts: certainty spreads faster than evidence. He was coming to understand that the charts might be the most dangerous thing he'd ever made.

He'd been at Vesper the day it went dark. Not by design — he ran a standing instrument on the inner gates, sampling the chatter, and Vesper had been a routine node until it wasn't, and so he'd happened to have his readers running at the exact moment a real silence met a frightened crowd, which for a man in his trade was like a stormchaser stumbling onto the eye. He'd watched the story assemble itself on the platform in real time. He still had the readings. In the first hour the dock had spoken in honest fragments — not syncing, probably a lag, nobody's saying, provisional, I heard, they think. He'd even contributed one himself, to a steward who'd asked: the family board's probably an indexing lag, provisional. By the time it came back to him off the market deck two hours later it had been Tarn entries not syncing, confirmed, in the mouth of a man who plainly believed he'd always known it. His own hedge, hardened and handed back. He'd stood there feeling the floor of his discipline tilt, because he'd just watched the mechanism work on himself, in two hours, in a sample size of one, and understood for the first time in his measuring life that he was not outside the thing he measured.

There'd been a courier. He kept coming back to her, though he didn't know why she stuck — the woman reading a false manifest at the wicket while the gate died, sharp-eyed, frightened in a way that was paying attention instead of panicking. He'd done a foolish thing on her account, the kind of thing his caution never let him do: he'd been pulling the gate's own records that morning, chasing a separate itch, and found the itch was real — auxiliary lines, return paths, legacy routes, every one tagged do not remove, and not one of them going where the register said it went. Records that described a network that no longer existed, or had never existed, he couldn't yet tell which. He hadn't been able to publish it; he couldn't even say what it was. So he'd copied the references out in his own cramped hand and sent them after her, a stranger heading the slow way north, because she was going where the records were wrong and he was a coward who could whisper a thing to one courier that he could not say to a room. None of these go where the records say they go, he'd written. Be careful what you trust. He had no idea if she'd lived. He thought about her more than was reasonable. She was the one person he'd told the true uncertain thing to, and he'd had to pick a person leaving forever to do it.

He went back through Cardinal to his reading-room — past the public feed-boards at the concourse mouth that had lately taken to repeating the same three confident words over and over, past the long quiet halls where analysts moved between reading-rooms like priests of a religion of caution, nodding to one another, saying nothing they couldn't footnote — and felt, as he always did here, the strange insulation of the place. The relay-masts hummed in the walls. The whole conversation of the inner worlds pooled and eddied around him, warm and lit and busy, while two worlds out the gates went dark one by one; and the town was so occupied with talking that it could not hear how much of what it said had become the same sentence. He did the work, because the work was the only place the panic in his chest resolved into something with axes and a scale.

His instruments at Cardinal pulled the whole inner chatter, and the whole inner chatter said sabotage. But Jun did not read what people said; he read how it moved.

On the left pane of his board he kept two lines. The black line marked gates that had actually failed. The red line marked where people were already blaming cutters.

In a real campaign the black would lead and the red would chase it. On his board the red led. At Maween, the red line arrived first. Jun checked the timestamp, disliked it, checked it again, and disliked it more. At Anneth, the same. At a dozen worlds he could set the rumor's first arrival against the first confirmed failure and find the talk running ahead — by days, sometimes; sometimes before there was any failure at all to attribute. He marked each one with a small careful tick. The ticks made a pattern of their own: the story arriving before the thing it explained.

The failures were real. He wasn't fool enough to doubt the dark gates. But the story about the failures was not tracking the failures. It was tracking itself — and he had the chart to prove it, and would not publish.

He sat with that, as he had for weeks. He wasn't sure — the signature was strong but not clean, three other mechanisms remained live, the sampling on the outer worlds was thin. That was the true reason and also the alibi.

The other reasons were uglier. He had watched what happened to the analyst who said the frightening counterintuitive thing and turned out to be even slightly wrong: they did not come back from it. And he had watched what happened to the analyst who hedged and was wrong: nothing. Hedging had built his career. On the one day it mattered, he would have nothing to offer but the data is consistent with several mechanisms.

So he refined the model instead. That was the joke he could appreciate even now, the mathematician's absurdity: faced with a human catastrophe his instruments could see and his courage could not name, he made the instruments better. He spent the night tightening the cascade model, improving the fit, adding a term for the qualifier-shedding he'd watched happen to his own provisional on the Vesper deck. It was elegant work. It was the best thing he'd ever done. He told himself he was strengthening it so that when he published it would be unassailable, and he half believed it, and the other half knew he was polishing a thing he intended never to use, the way a frightened man cleans a tool instead of going down to the cellar.

He saw the banner four days later, on the concourse, on his way to buy tea.

Varik's people had come to Cardinal — of course they had; it was the feed-hub, you took the inner worlds by taking the place they listened — and they'd hung their materials in the reading-quarter where the analysts walked, which Jun would later decide was either coincidence or the cruelest aim he'd ever been on the wrong end of. The banner was clean and handsome and printed in the confident block type of people who had stopped using qualifiers. It said, across the top: THE PATTERN IS CLEAR. And under it was a curve, pulled out of some quiet appendix and blown up to the width of a transit bay, built to be read across a concourse at a glance and believed before you'd stopped walking. It was drawn in a bold accusing red his own plots never used. His axes ran fine-lined and sober, in a muted blue, hedged round with footnotes in the margins — plots you had to lean in and weigh.

It was his curve.

He knew it the way you know your own hand. The shape of the propagation, the characteristic early steepening, the little shoulder where the talk crossed from the inner worlds to the outer — he'd derived that shoulder; it was his, it had taken him a season; he could have found it in a thousand. They'd taken it off his last published note, the careful one, the one so hedged it was practically apologizing — Preliminary: rate of propagation of interference-attribution exceeds rate of confirmable incidents; this divergence is consistent with several mechanisms including endogenous belief-cascade and should not, repeat not, be read as evidence of the attributed cause. Forty words of caveat around one curve. They had kept the curve. They had thrown away the forty words. And they had retitled it, and the title was the exact inversion of what the curve meant: his measurement of the belief outrunning the evidence was now hung in the reading-quarter as the proof the attack was spreading. The very steepness that proved the story was a story had become, stripped of its sentence, the proof the story was true.

He stood in front of it for a long time. People flowed around him. A young analyst he half-knew slowed and said, "Your propagation curve, isn't it? Good to see the work getting reach," and Jun said something, he never afterward knew what, and the young analyst went on, pleased for him, and that — the congratulation — was somehow the worst part, worse than the theft, the idea that this counted as his work getting reach. Reach was what you said when a careful thing had been made loud enough to stop being careful.

It had. That was the thing he had to stand there and take. His work had got more reach in four days on that banner than in twenty years in the journals. He had spent two decades learning to measure the gap between certainty and evidence, and the only time the world had ever wanted his findings was the moment someone deleted the gap and sold the certainty. He understood the mechanism perfectly. He had a term for it in his model, a beautiful term, the qualifier-shedding coefficient. And the mechanism had reached into his own desk drawer and done to his life's work exactly what he had watched it do to his provisional on the Vesper deck — filed off the maybe, handed back the confirmed — and he had not published the warning that might have inoculated against it, because he hadn't been sure, because being unsure out loud was how careers ended.

He was still standing there when the sub-prefect found him — the same man from the briefing, the folder-closer, who stopped at his shoulder and looked up at the banner with open approval. "There," he said. "You see? That's all anyone was asking for." He had no idea the curve was Jun's; he had no idea he was speaking to its author. He saw only the clean sentence he'd wanted, at last, hung where the whole Crossing could read it walking past. "Why didn't you just say this, Analyst?" Not cruelly — that was the unbearable part. He said it the way you'd thank a man for finally getting to the point. "Would've spared us all a great deal of fear." And he went off toward the reading-quarter, comforted, to be frightened of the right thing now, leaving Jun standing under the proof that the world had only ever wanted his evidence after someone had made it lie.

His career was, he supposed, going to be fine. That was the final cruelty. No one would ever know the curve was his to disown; his name wasn't on the banner, only his shape. He could go on hedging in briefing halls for another twenty years, safe and respected, while the thing he'd built convinced a frightened civilization to go looking for an enemy that wasn't there.

He didn't buy the tea.

He went back to the reading-room and sat in front of the beautiful, finished, unpublished model, and for the first time in twenty years the caution felt less like rigor and more like the thing it had always partly been, which was fear wearing rigor's coat. He had been so proud of refusing to claim more than he knew. He saw now that he had also refused to claim what he did know, and that the two refusals had worn the same face, and that he had let them, because the second one was comfortable and he had not wanted to look at it.

He thought about the courier again — the one he'd whispered the records to, the one person he'd trusted with an uncertain true thing because she was leaving and it was safe. He'd told himself that was generosity. It was the same coward's move as everything else: say the true thing only where it can't cost you. She was out there somewhere on the wrong-record routes, carrying his crumbs north, and he was in here with his finished model and his clean reputation, and the difference between them was that she had no instruments and was telling everyone the uncomfortable thing anyway, and getting thrown off docks for it, while he had the instruments and a hall that had literally asked him for the answer and he had said consistent with several mechanisms.

He did not, that night, become brave. He wanted to record that honestly, later, to himself: there was no scene, no decision, no rushing back to the briefing hall. He simply sat with the banner in his mind and the model on his desk and understood, for the first time, that he was not a neutral instrument that the cascade had unfortunately gotten hold of. He was part of it. He had measured the thing and refined the thing and, by his silence, fed the thing, and the precision he'd worshipped his whole life had turned out to be, ungoverned, just one more way of being part of the harm while keeping your hands clean.

He saved the model. He did not publish it. But he started, that night, a second document with no equations in it — no axes, no coefficients, nothing his discipline would recognize as work. He wrote it by hand, plainly. Short enough to be copied in an hour. Plain enough to lose few of its caveats on a long road. What the curve actually meant. How he knew. The briefing he'd hedged. The provisional he'd let harden on the Vesper deck. The forty words he'd buried the one warning inside.

He only knew that the hiding was over — not the caveats, but the hedge kept as an alibi, the qualifier mistaken for a place to stand where nothing could reach him. That had cost too much.