Chương 6: Episode 6 - Six Different Problems

By the end of the first week, Cassian had mapped the failure modes. Not fixed. Mapped. There was a meaningful difference, and he was keeping it clearly delineated in his own head because conflating the two was the kind of optimistic thinking that led to tactical overconfidence, and he had enough problems without adding that one.

Five drills across the first week — two morning sessions, two afternoon sessions, and the supplementary 0630 slot in the secondary bay that he'd booked because it was available and nobody else had claimed it. Nyx'ara had described this as "unhinged" and then shown up for it anyway. So had everyone else. He'd noted this. Each drill produced the same six problems in different configurations. Which was useful, in the clinical sense — the same way a pathology was useful.

It told you exactly what was wrong, with admirable consistency, even when what it told you was unpleasant. * Cirus froze when the plan diverged from the optimal path. Not always. Not randomly. Specifically, when the exercise parameters shifted in a way that invalidated the pre-planned approach, there was a window — three to eight seconds, depending on the severity of the change — during which RAZORWIND simply stopped. Not a malfunction. The frame's logs showed full operational status throughout.

The pilot had stopped, because he was running a real-time recalculation of every available path and hadn't yet found one that met his standards, and until he found one, he wasn't committing to any of them. After the second drill, Cassian had asked Cirus about it directly.

"I'm aware," Cirus said, before he'd finished framing the question.

"I know you're aware—""The recalculation time is a known variable. I am working on reducing it." He said it with the clipped precision of someone who had already run this internal meeting with himself and was not particularly interested in holding it again with an audience.

"The primary issue is that I'm evaluating against an optimality threshold that isn't appropriate for real-time adaptive conditions. I know this. Knowing it and correcting for it are two completely different problems." Cassian had looked at him. His crest was at trained neutral; his wings, folded against his back, were very still. He was waiting to be told this was adequate.

"Fair," Cassian said. He'd seemed faintly surprised. He made a note on his slate and didn't elaborate. * Nyx'ara improvised past the coordination point. Every time. Without exception. She would follow the plan until it presented her with an alternative that was better for SHADOWRUN specifically, and then she would take the alternative — which was always, always, the correct tactical choice for an interceptor-class frame operating independently. The problem was that SHADOWRUN was not operating independently.

He had tried explaining this in three different ways across three different debriefs.

"I understand the formation argument," she said after the third one.

"I understand it every time I make the call, and I make the call anyway." Her photodermis was still when she said it — the suppressed flat of someone choosing not to show how much the pattern frustrated her.

"That's the problem."

"I know it's the problem." She looked at him.

"The problem is my read is right. My positioning is right. What's wrong is that the formation isn't fast enough to use it. That's a design constraint, not a pilot error."

"Also correct," Cassian said.

"We're solving the design constraint." She looked at him.

"How long."

"I don't know yet." She was quiet for a moment.

"Give me time." He gave her time. He also gave her a flag count in the post-session data every day, because he was who he was. * Krell'va overcorrected. This was the logical consequence of the session-one debrief applied with High-Fang thoroughness. In the first evaluation, he'd broken ahead of the formation because the Howl-Node instinct outran the call. He'd heard the feedback, processed it, and implemented a correction. The correction was: do not break formation under any circumstances.

The result was that IRONJAW had become, across the subsequent drills, essentially a very large, very expensive, absolutely immovable object. Committed to formation discipline, just as a locked door was committed to security. Technically correct. Operationally limiting. Stolen content warning: this content belongs on Royal Road. Report any occurrences.

"You can move," Cassian told him, after the second drill where Krell'va had held position while the entire left flank of the exercise collapsed around him.

"I'm not asking you to stay perfectly still. I'm asking you to move with the formation."

"I am staying with the formation."

"The formation moved. You didn't." A pause. The ear tilt.

"The formation moved without a call."

"Sometimes the formation moves without a call. When that happens, follow the nearest squadmate." Krell'va considered this.

"Which one?"

"Whichever one is moving in the direction that makes tactical sense."

"They frequently move in different directions."

"I know. Pick the one that's going the right way." Krell'va looked at him with the expression of a man who had just been handed a problem requiring significant cognitive investment and was not complaining about this because complaining was not a concept he had much use for, but was communicating through other means that the problem was noted.

"I will attempt this," he said, with the air of someone who had accepted a legitimate tactical specification and intended to implement it to specification. He resecured the clasp on his battle-braid and turned back to the bay. The matter was apparently filed. * Korrath arrived calibrated and late.

Not a new problem, but a consistent one — the ritual sequence had a fixed duration, and the fixed duration and the exercise start time overlapped in a way that produced a two-to-four-minute window where HEARTHBREAKER was not in the bay, or was just arriving. The formation operated with a gap at the anchor point that everyone else quietly compensated for, with no one acknowledging it aloud. Nobody had said anything to Korrath directly. Cassian had noted this and decided it was currently the right call.

There was something the squad was working out about Korrath and his ritual. He left it in the file under: revisit at week three. * Ren-Thalen's problem was harder to observe because it operated below the surface of the exercise data. She was making decisions — clean, reasonable, technically correct decisions — that were consistently influenced by the wrong inputs. Not the tactical overlay. Not the formation position data. The ambient emotional state of whoever was nearest to her at the moment of decision.

He'd asked her about it.

"Yes," she said.

"I know." Her amber patterns were cycling at the slow frequency.

"The proximity signal is strong. I'm filtering for tactical relevance, but the filter is incomplete. In a close-quarters formation the emotional data arrives before the grid data, and I'm responding to what arrives first."

"Is the emotional data accurate?"

"It's accurate. It's not always the right thing to respond to." She paused.

"The distinction is the work." He had written: Ren-Thalen — making decisions from correct data at the wrong resolution. Not a tactical problem. A filtering problem. She already knows this. * He was still thinking about what to do with all of it when he ran the end-of-week analysis and found himself staring at his own entry. Cassian Vail. Over-plans. Cannot adapt when the plan meets reality. He'd written it himself, which meant he'd known it and had still spent the last five drills doing it anyway.

He had been building plans for six capable individuals rather than for one system. The plans had been correct about the individuals and wrong about everything else. He sat with that for a while. The tablet was face-up on the table. He looked at his own entry until he was done arguing with it. * At 2340 on the seventh day, sitting at B-7's common area table with six individual failure-mode documents open simultaneously, Cassian found the pattern.

He'd been staring at the six files and had the vague, itching sense of something he was almost seeing. The way you almost saw things in low light if you looked slightly to the side of them. He looked to the side of it. Cirus froze when the plan diverged. But Cirus also pre-calculated every contingency path to three decimal places before the exercise started, in the moments when everyone else was still reading the brief. Nyx'ara improvised past coordination.

But Nyx'ara could read a shifting tactical situation faster than anyone he'd ever trained with — faster than the sensor overlay updated. Krell'va moved too slowly within the formation. But Krell'va's threat assessment and physical positioning were already operating before anyone else's conscious analysis had started. Korrath arrived late. But when Korrath arrived, he arrived completely: no wasted movement, no hesitation, no performance ceiling that training pressure could push against.

Ren-Thalen made decisions from the squad's emotional state. But she was doing it four seconds before the tactical data caught up. Cassian overplanned and couldn't adapt. But he could see the shape of a six-person system well enough to be sitting at this table at 2340 and notice what it meant. He sat with the six failure modes laid out side by side and saw them clearly for the first time. None of them overlapped. Every single failure mode was distinct.

Cirus froze in the exact circumstances where Nyx'ara operated best. Krell'va over-held in the exact circumstances where Nyx'ara broke loose. Ren-Thalen's stress-tracking was filling gaps that Cassian's overplanning kept creating. Korrath's late-but-complete arrivals were filling an anchor position that nobody else in the squad was built for. They were failing in six completely different directions.

Which meant — theoretically — that if you could get six people who failed in six completely different directions to stop failing in them at the same time, you wouldn't just have a functional squad. You'd have a squad where each person's strength covered exactly the gap left by someone else's weakness. He opened a new document. He stared at the blank file for a moment. He was nineteen years old.

He'd been doing this since he was seventeen, which was not the same as being good at it, which was not the same as being right. He had been right before at significant cost and was not in the habit of trusting patterns just because they looked like patterns. COMPLEMENTARY FAILURE MODE HYPOTHESIS — REQUIRES FURTHER DATA. He saved it. Then he opened the drill protocol for the following morning and started from entry one. The hypothesis could wait for further data. The protocol couldn't wait for anything.

The lights were already out. He worked by tablet glow until 0100 and then went to sleep with the formation analysis still running in his head. The failure modes didn't overlap. He did not know yet whether that was good news or an interesting coincidence. He was going to find out.