205: The Platform Layer

The Tuesday call ran for two hours.

Wenqing, Ningxia, Cloudrift — the three-analyst configuration with Bai Yueran and me as guild leaders observing. The format was the same as the competition analysis calls: Wenqing's data first, then Ningxia's framework, then Cloudrift's cross-verification. We'd run this format through CW III and it worked because each person's knowledge domain covered a gap in the other two's coverage. Wenqing knew the formation data. Ningxia knew the competitive infrastructure. Cloudrift knew where those two things intersected in ways neither analyst had considered.

Ningxia's opening: "We started watching Lu Yifan's network in December because of an anomaly in the CW III bracket algorithm. The seeding calculation for the top-four seeds produced a result that was mathematically correct given the stated inputs, but the stated inputs had a minor discrepancy in the win-record weighting for the CW II championship." She shared the analysis — a chart showing the weighting parameters and the point where one value deviated from the specification by 0.4%. "The discrepancy was small — 0.4% — and the stated result was still Seed 1 to Black Dragon Guild. But the discrepancy was inconsistent with the algorithm's stated specification. Someone had adjusted a parameter."

Someone had adjusted the seeding parameter. The result had still been correct — we'd gotten Seed 1 — but the parameter had been touched. A finger on the scale that hadn't moved the scale, but had been placed there.

"Who has access to the algorithm inputs," I asked.

"The tournament administration platform. Eighteen individuals with administrator-level accounts. We ran the access log as publicly available through the platform's transparency policy." She showed the log. Eighteen names, eighteen accounts, eighteen rows of access history. "Twelve of the eighteen administrators have no observable connection to any guild or investor network. Four are affiliated with mid-tier guilds in coordination roles — no connection to Tianxia. Two are connected to the Lu Yifan investment network through employment records."

Two administrators connected to Lu Yifan's network.

"The parameter adjustment," Wenqing said. "It was a test. Not manipulation — a test. They adjusted the parameter, observed whether anyone noticed, then restored it. The final seeding was unaffected."

"They were testing their access," I said.

"Yes. Testing whether a parameter change produced observable reactions from the affected guilds. We didn't react in December because we didn't notice until Ningxia identified it in January." He paused. "Which suggests they concluded the test was not observed. Their assumption is that their access is undetected."

So Lu Yifan's network had tested access to the seeding algorithm in September, before CW III began. They hadn't manipulated the result — they'd established that they could. A reconnaissance move. The kind you made when you were planning something that required infrastructure access but weren't ready to commit.

The platform infrastructure layer.

Wenqing: "The pattern of platform tests since CW III: four separate access events across three platform functions. The seeding algorithm, the performance verification module, the ladder ranking calculation, and the session validity checker. All four tests restored to original state after the test. None produced observable effects on any outcome."

Four tests. The picture was forming. Not chaotic probing — systematic. The four functions together covered the complete landscape of how competitive results were recorded, verified, and ranked. Someone was mapping the full capability.

"What does the full manipulation capability look like if all four functions are active," I asked.

Ningxia: "At full activation, the four platform functions could: artificially lower a guild's ladder ranking, flag a guild's sessions as invalid and remove them from the competitive record, adjust the seeding algorithm to prevent specific seeding outcomes, and manipulate the performance verification to affect the class capability acknowledgment."

The performance verification module.

The Heaven-Severing class's designation in the competitive record was verified by the performance module. If Lu Yifan's network invalidated the performance verification, the class designation in the achievement board could be flagged as unverifiable. Not deleted — flagged. Which was almost worse. A flag created ambiguity, and ambiguity in competitive records created grounds for challenges that could tie the guild up in procedural disputes for months.

That was the specific point of leverage against our guild that didn't exist against any other guild on the server. We were the only guild with a Heritage class record in the performance verification system. The only target that had something unique enough to be worth targeting.

"They're targeting the class record," I said.

Ningxia: "That's our read. The class activation is the competitive advantage that's least replicable — Wang Jian can improve his tactical modeling, Bai Yueran can develop her formation, other guilds can recruit better players. But they can't replicate the Pioneer's Path quest chain. The class record is the permanent differentiator."

The thing that made the guild's position unassailable was also the thing that made it the specific target.

Wenqing: "Wenqing's archive. The full documentation exists independently of the platform record. If the platform record is manipulated, the archive is the counter-evidence."

He'd been building the counter-evidence before knowing it was counter-evidence. Three years of documentation that was now, retroactively, exactly what was needed.

"What's the timeline," I said. "How long before they move from testing to activation."

Ningxia: "Unknown. The test pattern is accelerating — the first test was in September, the second in November, the third in February, the fourth in March. The interval is shortening. If the pattern continues, the next event is May or June."

May or June.

Thesis defense was June 22.

The calendar had a certain irony to it. The two significant things arriving in the same month window, requiring attention in opposite directions simultaneously. I'd survived worse timing in the previous timeline. The difference was that now I had people who watched the platform and people who watched the formation and people who watched the formation's opponents, and they were all talking to each other.

"We need a parallel record outside Wenqing's archive," I said. "A public-facing confirmation of the class record and all session data before May."

Wenqing: "I can format a public certification submission to the Tianlong server's official historical record database. It's a different system from the tournament administration platform — separate infrastructure, separate administrators, no connection to the Lu Yifan network's access map. If the certification goes in before the manipulation, the historical record pre-dates any platform change."

"How long does the certification submission take."

"Twenty-four hours. I'll submit tonight."

Bai Yueran, who had been listening through the call without interjecting: "Ningxia can co-certify on behalf of MoonShadow's competitive record documentation. A second-guild co-certification strengthens the authenticity of the filing."

Ningxia: "Already drafting the co-certification. Sending it to Wenqing in twenty minutes."

Wenqing submitted the certification at eleven PM.

I watched the submission notification arrive in the guild communication system and then set the phone down on the desk. Chapter 4 was still open on the laptop. The abstract notation looked the same as it had before the call. The work was still the work.

***

Wanqing was at the Thursday bench when I told her the details.

She listened to the full sequence — the September parameter test, the four platform functions, the class record as the leverage point, the certification submission. The April campus had the established-spring density now, the full canopy a different quality of light from the young-leaf week. Darker, more certain of itself.

"He built the archive as documentation," she said. "And the documentation is now the defence against the manipulation."

"Yes."

"He didn't plan it that way."

"No. He documented because it was worth documenting. The defensive value is a consequence of the quality of the documentation." I looked at the established maple. "The same way the workshop model's referral network section became the evidence base for the Q2 revenue projection. You built it because it was worth building. The application followed."

She turned a page. "The defence against platform manipulation," she said, "is having been so rigorous that the record existed before the manipulation."

"Yes."

She was quiet for a moment. The bench. The April campus. The sound of a seminar room door opening and closing somewhere in the adjacent building. "And if they try to manipulate anyway — after the certification."

"Then the pre-dating is evidence of manipulation. The record that was submitted before the manipulation contradicts any retroactive platform change. The manipulation becomes self-documenting."

She turned another page. "Does Wang Jian know," she said.

"About Lu Yifan's platform strategy. Unknown. Wang Jian has been the front — the coalition's competitive behavior was the visible layer. Whether he knows the full scope of what's behind him, I don't know." I looked at the maple. The full-spring canopy. "He's smart enough to suspect there's a floor he can't see. Whether he's asked is a different question."

"But you knew."

"I knew the name was coming. Not the specific form."

She looked at the maple. "The previous timeline's version — what happened to the guild."

"The guild survived it. At cost. Two members lost their ladder rankings to the performance verification manipulation — they couldn't compete in the bracket tier for six months. The class record wasn't touched in the previous timeline because the previous timeline's guild hadn't reached the Heaven-Severing class before the disruption phase."

She turned the page. "This timeline's guild is better positioned."

"Yes. The certification is submitted. Wenqing's archive is pre-dated. The alliance has been watching since December. We're not in the same position the previous timeline's guild was in."

She looked at the problem set. Not reading it — looking at it. The habit of a person who thinks better with something in front of their hands. "Is there anything else he can do," she said. "Lu Yifan."

In the previous timeline's disruption phase, there had been a second vector: the personal layer. Targeting guild members' IRL situations. Creating pressure outside the game that affected the game — the kind of pressure that was harder to document because it happened in the grey space between professional life and competitive play, where the connection was real but deniable.

I hadn't told Wanqing about that vector.

"Possibly," I said.

She looked at me. The direct look. "Possibly."

"The platform layer is the one Wenqing and Ningxia can track. There may be other approaches."

She held my gaze. Outside the campus was doing what the campus always did in spring — going about its business without caring what was happening at any particular bench. "Tell me," she said, "when you know."

"I will."

She went back to the problem set.

I looked at the April maple. The certification had been submitted. The archive was protected. The alliance was watching. The question was whether that was enough, and what Lu Yifan's network's patience level was, and whether the personal vector was already in motion.

In the previous timeline I'd found out it was in motion when it was already too late to be ahead of it.

I watched the maple and thought about patience.

The April campus continued around us. A student with a folder crossed the path below, stopped to check something on his phone, continued. A bicycle on the road at the edge of the grounds. The particular background noise of a campus that was present and unconcerned. None of it knew about Fengrui Capital or the seeding algorithm or the parameter that had been touched and restored. The campus just moved through its Tuesday the way it always moved through its Tuesday. The indifference was useful — it reminded you that the things that mattered were not legible from the outside as things that mattered. They looked like an ordinary afternoon.

The personal vector in the previous timeline had arrived through a guild member's employer — a false report, a manufactured performance issue, the kind of institutional pressure that was nearly impossible to counter quickly because institutions moved slowly. The counter in the previous timeline had taken three weeks to assemble and had arrived too late to prevent the six-month ranking disqualification.

The personal vector's logic was simple: if you can't attack the guild directly, attack the people in it through the channels those people already exist in. The channels they'd spent years building — their careers, their relationships, their professional networks. The channels where the guild was visible and the manipulation could arrive in a plausible guise.

I thought about who in the current guild had those kinds of channels. Who had careers that could be targeted. Who was reachable.

The guild's members were mostly players — students, young professionals, people in the early years of building. The channels were there, if someone wanted to find them.

The alliance had channels too.

I thought about MoonShadow. Bai Yueran's research network, Ningxia's platform connections, the analyst layer that had been building the investor network map since December. All of those were professional channels, attached to real people with real careers. If Lu Yifan's network had noticed the counter-surveillance — and it was possible, the counter-surveillance had been active for four months — the alliance's channels were reachable.

I needed to say this to Bai Yueran.

I picked up the phone.

"I'll message her tonight," I said. Not to Wanqing specifically — to the bench, to the decision that had just been made.

Wanqing, without looking up: "Good."

The problem set. The April maple. The platform documentation submitted. The personal vector not yet arrived, but the warning spoken.

That was the difference between this timeline and the previous one: speaking the warning before the arrival.