Wittaric's reply arrived on the fifth day.
Not through official courier — through Adalric, who had apparently found a merchant going south and had tucked the letter in with goods bound for Marcianopolis with the specific pragmatism of someone who knew the official courier system was slow and unreliable and had spent forty years finding faster routes.
Abster read it in the headquarters anteroom while Flavius managed the morning's incoming reports around him.
Wittaric had answered the question about coordination signals. He had also written something Abster hadn't asked for, which turned out to be the more important part.
The signals first.
Three of them, as Abster had suspected — one from the screen commander to the jaw, one from the jaw commander to the weight, one emergency signal that meant abort the encirclement, extract immediately. The first two were visual: specific patterns of spear movement against the sky, readable at distance but not obvious to an observer who didn't know what to look for. The third was audible: a specific rhythm on a drum that Wittaric described as the sound of the retreat from a bad crossing, which Abster understood was a memory rather than a military specification.
He copied these carefully and set them aside for Marcella.
Then he read the rest.
You ask why the screen works against Roman infantry when it does not work against our own cavalry. The answer is not the screen itself. The screen exploits the Roman understanding of what a retreat means.
When Gothic cavalry retreats, every Roman soldier who sees it understands the same thing: we are winning, they are running, pursue. This is what Roman military culture teaches. A retreating enemy is a breaking enemy. A breaking enemy must be chased.
We do not teach this. When our cavalry retreats, our other cavalry asks: why are they retreating? What did they see? Where are they going? A Gothic rider who retreats is communicating something. A Roman soldier who retreats is failing.
This is not a small difference. It is the difference between how our peoples understand movement.
Abster stopped reading and looked at the ceiling.
The screen exploits the Roman understanding of what a retreat means.
He had been thinking of this as a tactical problem — how to train soldiers to hold against a feint. What Wittaric was describing was something larger: two different civilizations with two different relationships to the same physical act. Roman culture read retreat as failure and therefore pursued it. Gothic culture read retreat as information and therefore interrogated it.
He kept reading.
My grandfather's people came from east of the Carpathians. The Huns drove his grandfather's people west. His grandfather's grandfather was driven west before that, I think, though we do not have records that go back far. We have been moving west for three generations.
When we arrived at the Roman river, we were not trying to take Rome. We were trying to stop moving. We had been moving for so long that stopping felt like dying, and we came to Rome's river and asked to stop, and Rome said yes and then no and yes again and took our swords and our grain and our children.
I tell you this not for your sympathy. I tell you this because you should understand what your soldiers face when they face mine. They face men who have been moving for a hundred years and are very tired of it and are not going to move again. Men who are tired in that specific way fight differently from men who have other options. They do not calculate whether to hold. They hold because holding is the only thing that means anything.
This is why the jaw does not break even when the Roman counterattack is strong. It is not superior tactics. It is different stakes.
Abster set the letter down.
He sat with it for a long moment.
Then he picked up his stylus and started writing.
He had been trying to write the answer for Betto for three days.
The answer Sebastianus wanted: a framework for managed trust. Something that addressed Betto's argument — fear is consistent, trust is not, seventeen men dead in their beds — without simply asserting that trust was better. Because trust wasn't simply better. Betto wasn't wrong. He was incomplete.
What Wittaric's letter had given him was the piece he'd been missing.
The seventeen men who had died in Betto's predecessor's command had died because the commander had extended trust as though it were a single act rather than a system. He had trusted one person and that person had proved untrustworthy and the trust had no architecture to limit the damage.
Trust as a single act was fragile. Trust as a system was different.
He wrote for two hours without stopping. What emerged was rough — not doctrine, not a manual, something more like thinking out loud in formal language. He called it, in his head, the Framework of Managed Trust, though he didn't write that title on it because a document with that title written by an optio would be dismissed before anyone read the first sentence.
The principles:
Trust is extended in layers, each layer conditional on the previous layer's demonstrated reliability. You do not trust a man with battle positions until he has demonstrated reliability in smaller things. You do not trust a community with your plans until the community has demonstrated consistent behavior over time.
Trust is compartmentalized by exposure. A trusted source who proves untrustworthy damages only the intelligence that source had access to, not the entire network, if the network is properly compartmentalized.
Trust is monitored rather than assumed. A relationship of trust is not a relationship of blind faith. It includes ongoing verification, which is not suspicion — it is the maintenance of the system.
Trust is scaled to context. The level of trust appropriate in a stable community with six months of established relationships is different from the level appropriate in a community one week after a crisis.
Fear, as a governance tool, is cheaper in the short term and more expensive in the long term. It does not require the investment of trust-building. It also does not produce the benefits of trust: cooperation, information-sharing, active support rather than passive compliance.
He read this back.
It wasn't finished. It had gaps he couldn't fill yet — the question of how you built verification mechanisms, the question of what you did when trust broke down at scale, the question of whether the managed trust system was itself fragile in ways that fear-based systems weren't.
But it was a start. And it was, more importantly, an answer that didn't claim trust was simply superior. It acknowledged the cost. It acknowledged Betto's seventeen dead men. It said: here is a system that accounts for those deaths while producing better outcomes than the alternative over time.
It was not clean. Real governance never was.
He set it aside for Sebastianus.
That afternoon he walked the city again.
Not planning to. Just walked, the way he'd learned to walk cities — letting the actual thing in front of him replace the version he'd been carrying in his head.
The forum near the basilica had changed since his first day.
There was a line outside the bishop's building — not worshippers, not the orderly kind of line. A grain line. People with containers, waiting. A deacon was managing it with the focused efficiency of someone who had been doing this for longer than he'd expected to be doing it.
Abster stopped and watched.
The line was mixed: Roman citizens, Gothic refugees, people who were neither cleanly and had been living in that ambiguity long enough to have developed their own category. The grain was being distributed from sacks inside the doorway — not military ration sacks, smaller, the kind that came from private stores. The bishop's own grain. Or grain the bishop had collected from somewhere.
A woman near the front of the line — Roman, older, the kind of worn that came from years of managing a household on insufficient resources — was arguing with the deacon in the specific low-intensity way of someone who had been doing this long enough to know which arguments moved things. Behind her a Gothic man in his thirties was watching the argument with the carefully controlled expression of someone who had learned that visible impatience in Roman lines produced bad results.
Two people waiting for the same resource, each calculating how much frustration they were allowed to show.
He watched for a while.
There was a child — seven, maybe eight, Gothic by her coloring, with the specific alert-and-watchful quality that Abster had learned to read as this child has been managing her own safety for longer than children should have to. She was not in the line. She was watching it from a corner, the way Amalasuntha had watched the drills — trying to understand something from the outside.
He thought about Amalasuntha.
He thought about the carved bone eagle, still in his kit.
He should go back. He should be going back. Six weeks and he'd been here five days. Twenty-three days left.
He kept watching the grain line until the deacon reached the Gothic man, who received his allocation and walked away with the specific gait of someone who had been given less than he needed but more than he'd feared.
The civilization being fought over, he thought again. But more specifically this time: The civilization that is already losing the fight in ways that don't show up in any military report.
He went back to the headquarters and found Sebastianus.
"The bishop's grain," Abster said.
Sebastianus looked up.
"His building has a line outside it. He's distributing grain to civilians — Roman and Gothic mixed. More demand than supply. I watched it for twenty minutes." He sat. "The official grain distribution isn't reaching the civilian population adequately. The bishop is filling the gap with his own resources, which will run out."
"I know," Sebastianus said.
"The grain prefect's office—"
"I know." He turned from the map. "Dextrus is managing within a system that is structurally extracting from the supply chain at multiple points. He can't fix the structure. He can optimize within it." He paused. "The bishop's grain line will run out in approximately three weeks, by my estimate. At which point the civilian population of Marcianopolis — which includes a significant number of Gothic refugees — will have food insecurity that the military situation will not contain."
"What happens then."
"Civil unrest. Possibly riots. The Gothic refugees become associated with the scarcity in the minds of the Roman population who are also experiencing the scarcity. The association produces violence. The violence produces a military response that undoes six months of careful relationship-building." He looked at Abster. "This is the layer of the problem that isn't in any of Sebastianus's tactical documents. It's not a military problem. It's a supply chain problem that produces military consequences."
"And Dextrus."
"Dextrus is an honest man in a dishonest system. He can give me three weeks' warning. He can't give me grain." He went back to the map. "This is why the grain contracts matter. And why the church matters. And why Treverius matters even though he's complicated." He paused. "The problem isn't solvable through military reform alone. You're going to have to start thinking about that."
Abster was quiet for a moment.
"Wittaric wrote that the Goths have been moving west for three generations," he said. "Driven by the Huns. His grandfather's grandfather was displaced from the eastern steppe. They came to Rome's river trying to stop moving." He paused. "The grain line in front of the bishop's building is the end of a three-generation migration. That's what we're managing."
Sebastianus looked at him. "Yes."
"The answer for Betto," Abster said, "is incomplete if it only addresses the military relationship. If the civilian population collapses, the military relationships don't matter."
"Also yes." He sat. "Welcome to Arc Two of the problem. Which is also Arc Three and Arc Four, simultaneously." He looked at the maps on the wall — the frontier lines, the Gothic positions, the Roman defensive dispositions. "What did Wittaric say about the signals?"
Abster was crossing the forum toward the waystation that evening when someone said, behind him, loud enough to carry: "There goes the Gothic optio."
He kept walking. Didn't turn.
It had been said without obvious hostility — the casual observation of two soldiers having a conversation, the kind of comment that gets made when it's not intended to land and sometimes does anyway. He couldn't tell from the voice whether they knew he'd heard.
He walked back to the waystation and sat alone for a moment.
The Gothic optio.
Not a name. A category. A way of filing him that made his methods the thing he was rather than something he'd chosen.
He thought about Betto's version of the same filing: an ambitious ranker who operates outside procedure. He thought about what Treverius had said: two narratives competing in other people's minds. He thought about how long it had taken him to build the trust in the XII's sector, and how quickly the phrase Gothic optio could undercut it in the minds of men who hadn't been there.
He could not control the phrase. He could not un-say it. He could not, probably, stop it from spreading.
He wrote it in the vellum that night.
Then he wrote: Betto is right that trust has failure modes. He is wrong that fear is the answer. The answer is managed trust, which requires architecture I am only beginning to understand. The document is not finished.
The grain line at the bishop's building has three weeks left. After that: civil unrest, probably. The Gothic refugees and the Roman civilians will both be hungry and they will blame each other and the military situation will not contain it.
Five hundred and forty-four days.
The Gothic optio. I'll carry it. But I need to be more careful about what I say in rooms where I don't control the framing.
He folded it. Kept it.
Two days later, Sebastianus came to find him at the practice ground where he'd been working with Modestus.
He arrived between exercise sets — deliberately, Abster thought, so that Modestus would see a tribune seeking out an optio rather than the reverse.
"I read the trust framework document," Sebastianus said.
Modestus found somewhere else to be. Abster waited.
"It's incomplete," Sebastianus said.
"I know. I haven't solved the verification mechanisms yet."
"The verification mechanisms are a separate document. What you have is the philosophical core." He paused. "Betto read it this morning."
Abster looked at him. "You gave it to Betto."
"I showed it to Betto. Without your name on it. As a potential policy framework for Gothic sector management." He met Abster's eyes. "He said it was theoretical and unproven at scale."
"He's right."
"He also said, and this is the part you should know, that the framework addressed his objections." He paused. "Not that it resolved them. Addressed. He said: this is at least thinking about the problem correctly, rather than pretending the problem doesn't exist."
Abster sat with this.
Not a conversion. Not Betto changing his position. Just: acknowledgment that the argument was being made honestly, by someone who had understood the failure mode he was worried about.
It was enough. For now. More than he'd expected.
"He also said," Sebastianus continued, "that he wants to know who wrote it."
"What did you tell him."
"I told him the relevant question was whether it was right, not who wrote it." He paused. "That will hold for approximately three days. You should be aware."
Abster nodded.
"There's something else," Sebastianus said. He looked at the practice ground rather than at Abster — the posture of a man about to say something he'd been deciding whether to say. "The court in Constantinople. Lucius Bassus has been in correspondence with someone in Valens's personal secretariat. Not about you specifically — about the tactical reforms. The framing in the correspondence, from what Treverius was able to determine, characterizes the reforms as..." He paused. "As weakening military discipline by introducing excessive consideration for barbarian sensibilities."
"That's Betto's argument."
"It's Betto's argument having reached court, which is a different thing." He turned. "Valens is not an unreachable emperor. He's a man under considerable political pressure who makes decisions based on information provided by people with competing interests. The information reaching him about these reforms is currently unfavorable." He paused. "I have limited court access. I have spent most of it on you already."
It landed clearly: Sebastianus was telling him that he was exposed, that the court was a problem he couldn't solve on Sebastianus's political capital, and that Sebastianus's own position was not as secure as it appeared.
"How vulnerable are you," Abster said.
"Not immediately." He looked at the map he'd brought — habit, always a map. "I have a record of tactical successes. I have the eastern field army's confidence, which is worth something. I have Gratian's court in the west, which corresponds with Valens's court and sometimes provides useful pressure." He paused. "But I have been arguing for these reforms for three years. Every year I argue for them and nothing happens is another year of my argument being ignored, which is its own kind of evidence against my judgment."
"If the reforms work—"
"If the reforms work conspicuously, before Valens has committed to a position against them, the political situation changes." He folded the map. "That is why I need you to move faster than feels comfortable. Not recklessly. But the window is not infinite."
He walked away.
Abster stood on the practice ground and thought about three-year arguments that were slowly losing credibility, and about Lucius Bassus in the imperial secretariat, and about the grain line outside the bishop's building that had three weeks left.
Five hundred and forty-two days.
The practice ground was empty now except for him. The sun was low, the late-afternoon light the kind that made the city look older and more beautiful than it probably deserved.
He started the footwork forms alone, working through them slowly, because it was something his body knew how to do and because the repetition was clarifying. Screen. Jaw. Weight. Hold the line. Don't pursue the retreat. Move when the jaw shows.
He thought about three generations of Wittaric's people moving west.
He thought about a child watching a grain line from a corner.
He thought about what it meant to be building something you knew might not be enough, and doing it anyway because stopping was worse than failing.
Do the next thing.
He finished the forms. Went back to work.