Mao Yingjie arrived at one fifty-eight.
She was twenty-nine, B-rank advancing, and she walked into Han Weiwei's fourteenth-floor waiting area with the specific alert quality of someone who had spent four years operating in isolation and had developed, as a consequence, an extremely accurate read on unfamiliar environments. She clocked the administrative team, the floor layout, the exits — not obviously, the way a combat-type did it — but he noticed the sequence of her attention moving through the room before it reached him.
She found him standing by the window.
"You're shorter than I expected," she said.
"How did you form an expectation?" he said.
"The task completion data." She looked him over once. "High-yield, consistent, multi-domain. I expected someone who looked more impressive."
"You've been running four-year-old expectations," he said. "I was unimpressive for most of them."
A short pause. She decided something about him in that pause — he could tell from the slight shift in her attention quality, something becoming less cautious — and then she said, "Fair," and looked at the door to Han Weiwei's office.
He had been watching her for the forty seconds she'd been in the room and building a more accurate picture than Tang Mingxiu's profile had given him.
Mao Yingjie was compact and quick-moving, with the cultivation signature of someone who had been developing under unusual parameters for years — her field was layered in a way that suggested multiple ability threads developing in parallel rather than a single dominant expression. That was different from his own architecture, where the probability field ran as a passive ambient layer and the cultivation sat beneath it like a foundation. Her mechanism appeared to have built her something more like a web.
He updated his model of her and stored it.
"She's going to be direct," he said. "Answer directly. She's not a threat."
"I know she's not a threat," Mao Yingjie said. "She's a Bureau researcher who's been working on this from the outside. Her files don't indicate hostile intent — she's intellectually curious, not politically motivated." A pause. "I read her publication record last night. She's good."
"Yes," he said. "She is."
The door opened. Han Weiwei came out to meet them rather than waiting inside, which he noted as a choice — the meeting-on-equal-ground posture, deliberate from someone who had probably spent the morning deciding how to set the dynamic.
"Mao Yingjie," she said. "Thank you for coming."
Mao Yingjie looked at her for a moment. Then: "You found the mechanism from the outside."
"I found the field signature data and the distribution pattern," Han Weiwei said. "The mechanism itself is what I need the session for."
"All right," Mao Yingjie said. "Let's go in."
---
The session ran three hours.
Han Weiwei had seven prepared questions. She got through three of them before the conversation had moved far enough beyond their structure that she set her notes aside and just asked whatever the current moment demanded.
The first thing Chen confirmed: his mechanism was real, it had deployed approximately eighteen months after his zero-assessment, and its primary function was task generation with LP accumulation and expenditure as the core loop.
He didn't say the system called it a Luck System. He said: "Probability field generation and management, task-based progression, an internal shop economy." Accurate, partial.
"The probability field is what I've been reading in the field audit data," Han Weiwei said.
"Yes."
"It runs passively."
"Always. It doesn't require activation. It's ambient."
She was making notes. "What does it feel like from the inside?"
He thought about how to answer this accurately.
"Like background weather," he said. "At Level 1 it was noticeable — you'd feel the small coincidences operate. The door being unlocked at the right moment. A relevant person glancing the other way. Over time it became ambient enough that I stopped registering it as separate from normal circumstances." He paused. "Which is probably the correct calibration. Relying on it as a separate tool would create a dependency that doesn't scale well."
"You're not relying on it," she said.
"I use it as a foundation," he said. "Not a substitute."
Mao Yingjie said: "Mine is different."
They both looked at her.
She had been listening carefully since the session started — not passively, but with the quality of someone cross-referencing what she heard against a large internal model. Now she leaned forward.
"My mechanism doesn't generate a probability field," she said. "It generates a pattern recognition layer. I see connection opportunities that other people don't register — not through ability, through an attention filter my mechanism put in place." She paused. "I can see, at any point, approximately six to eight nodes in my social and professional network where a specific action would generate disproportionate positive outcomes. The mechanism flags them. The task is usually: connect these two people, facilitate this exchange, introduce this piece of information here." A pause. "I've been bad at completing them because I kept waiting until the conditions were perfect."
Han Weiwei looked at her. "The conditions?"
"When you see the opportunities very clearly, it's easy to believe that waiting for a slightly better moment will produce a slightly better outcome," Mao Yingjie said. "Four years of that habit. I've left a lot of LP on the table."
"How much?" Han Weiwei asked.
"My mechanism's estimate is that I've operated at about thirty percent of my optimal rate," Mao Yingjie said, with the flat precision of someone who had run this number and found it difficult.
Chen said nothing. Mao Yingjie had named her own failure clearly and didn't need it commented on.
Han Weiwei looked at Chen. "Yours?"
"Higher than optimal at the beginning, when I didn't have full context. Lower than optimal during a specific three-week period at month seven when I overtasked and burned the probability field's efficiency window." He paused. "Approximately eighty-five percent of theoretical optimal, aggregated. I've been running closer to ninety-two for the last six months."
"You track it," Han Weiwei said.
"The mechanism tracks it. I read what it generates."
She looked between them for a moment. He could read the quality of her processing — the pattern analyst at full operation, taking in two live data points for a theory she had been building from external signatures. Her spatial ability's field had shifted slightly, the compressed-potential signature running at a higher register than its baseline. Not combat activation. Cognitive activation, which was apparently how her particular ability expressed intense focus.
He found this specific.
"The collaborative architecture note," Han Weiwei said. "You mentioned it. Both mechanisms generated it."
"Mine appeared two weeks ago," Mao Yingjie said. "His appeared four days ago." She looked at Chen. "Why the gap?"
"Your mechanism may have identified the network opportunity earlier because your pattern-recognition layer processes social connection nodes more directly than my mechanism does," he said. "The Luck System's primary metric is probability reduction. It would identify the collaborative network benefit after enough individual-level data had accumulated to make the comparison accurate." He paused. "The fourteen-day gap suggests my mechanism ran more data before flagging the opportunity."
"Conservative," Mao Yingjie said.
"Precise," he said.
A pause. She accepted the correction.
Han Weiwei set her pen down. "The third node."
They looked at her.
"Your mechanisms both generated a note about a three-node collaborative network," she said. "Two nodes are in this room. The note described the third as already identified or accessible." She met his eyes. "You came here for the research collaboration. Your mechanism sent you here. The three-node architecture requires someone in this building."
He had wondered when she would get there.
"The mechanism hasn't formally designated you as the third node," he said. "The collaborative architecture task specified initializing the network with Mao Yingjie and — in the framing used — the researcher whose work connects the mechanism to the external world."
Han Weiwei was quiet for a moment. "That's me."
"The task rewards for establishing this network have been structured around the research collaboration being real," he said. "Not around the proximity to the Bureau's database. The collaboration first, the connection second."
"But the connection is what activates the third node," she said.
"Yes."
She looked at him. At Mao Yingjie. Back to him. "I don't have a mechanism."
"No," he said. "Your role in the collaborative architecture may not require one. The mechanism's network design appears to function differently — some nodes contribute the internal processing, some contribute the external validation framework. Your research generates the external map that helps the network understand what the mechanisms are doing." He paused. "This is inference. I don't have full data on how the three-node architecture functions."
"Neither do I," she said. "Yet."
A pause.
"Yet," Mao Yingjie said quietly.
---
The session ended at five.
Han Weiwei walked them to the elevator. Mao Yingjie went down first — she had a rail connection to catch. Han Weiwei stayed at the elevator bank after the doors closed.
"Thursday morning," she said. "Seven AM. I want to run the field signature data against the session notes before it fades." She looked at him. "There are things you didn't say."
"Yes," he said.
"I'm not pushing on them."
"I know."
"I'm noting that I noticed," she said. "For the sake of accuracy."
He held her gaze. The spatial ability's field ran at its standard baseline again — the cognitive activation had settled. The elevator bank's ambient field was neutral, the building's standard register.
"The parameters I set on what I confirm are mine to maintain," he said. "The research benefits from what I can give within those parameters. Both things remain true."
"Both things remain true," she said. "I know." A small pause. "You're careful about this."
"There's specific reason to be careful," he said.
"I believe you." She looked at him — not the researcher's assessment gaze, something slightly different. "I've been building this research for two years. The pattern data was compelling when I had ten cases. With your mechanism as a confirmed active case and Mao Yingjie's as a second data point, the research is now—"
"Different," he said.
"Very different." She met his eyes. "I won't misuse what you've given me."
He thought about this honestly — whether that was a calculation he was in a position to make accurately. Whether the research framework as he'd described it, without the full system interface, was containable in a way that served both their purposes without creating the kind of exposure the system architecture was clearly designed to avoid.
"Thursday at seven," he said.
She nodded once. The elevator arrived. He stepped in.
The doors closed.
He looked at his reflection in the elevator's polished surface. The city was visible through the building's window panels as the elevator descended — its standard late-afternoon configuration, the Central District running its after-work cycle.
```
[SIDE TASK: THREE-NODE NETWORK INITIALIZATION — STAGE 2 COMPLETE]
[HAN WEIWEI: NETWORK AWARENESS ESTABLISHED]
[CITY RECIPIENT NETWORK: 3 OF 3 NODES AWARE AND IN COLLABORATIVE CONTACT]
[LP AWARDED: 14,600]
[TOTAL LP: 177,300]
[NOTE: NETWORK-LEVEL TASK GENERATION WILL INITIALIZE WITHIN 72 HOURS OF FINAL NODE CONFIRMATION. ANTICIPATE TASK ARCHITECTURE DIFFERENT FROM CURRENT INDIVIDUAL TASK STRUCTURE.]
[REMAINING TO LEVEL 4: 22,700]
[REVISED ESTIMATE: 7–10 DAYS AT CURRENT RATE.]
```
He read it twice.
Seven to ten days.
He had been running the Level 3 probability field for months. It had become ambient, unremarkable, part of the background register of how the world operated around him. Level 4 was a category shift: near-deterministic outcomes, opposition beginning to feel like they were fighting something larger than one person.
Ten days.
---
He was two blocks from the Bureau building when the system generated the new task.
Not a side task. The formatting was different — cleaner, the header block running longer than standard.
```
[NETWORK TASK — FIRST GENERATION]
[TASK TYPE: COORDINATED APPROACH]
[SUMMARY: MAO YINGJIE HAS IDENTIFIED A HIGH-VALUE INFORMATION NODE IN HER WESTERN DISTRICT NETWORK. A TIER-2 FACTION INTELLIGENCE ASSET HAS BEEN GENERATING UNAUTHORIZED SURVEILLANCE REPORTS ON INDEPENDENT PRACTITIONERS IN THREE DISTRICTS INCLUDING THE WESTERN DISTRICT. THE ASSET'S OPERATING LOG CONTAINS NAMES AND PATTERN DATA THAT WOULD SIGNIFICANTLY ADVANCE HAN WEIWEI'S DISTRIBUTION RESEARCH TIMELINE.]
[TASK: ACQUIRE AND ROUTE THE OPERATING LOG TO HAN WEIWEI'S RESEARCH FILE THROUGH UNTRACEABLE CHANNELS.]
[COORDINATION: MAO YINGJIE HAS THE SOCIAL CONNECTION. YOU HAVE THE TECHNICAL ACCESS ARCHITECTURE. HAN WEIWEI HAS THE DESTINATION FRAMEWORK.]
[REWARD: 18,400 LP — DISTRIBUTED EQUALLY ACROSS THREE NODES]
[ADDITIONAL REWARD: HAN WEIWEI RESEARCH TIMELINE: ADVANCED BY APPROXIMATELY 6 WEEKS]
```
He stopped walking on the sidewalk.
18,400 LP total. 6,133 LP per node. Not enormous by his individual side task standards at S-rank — but this was the first network task. The first one the system had generated.
And it advanced Han Weiwei's research timeline by six weeks.
He read the task structure twice, then sent a message to Mao Yingjie: *Check your mechanism. New task format.*
Her response came in ninety seconds: *I see it. The faction intelligence asset — I know which one. I can make the connection Tuesday.*
He sent: *Tuesday is correct. I'll handle the technical routing.*
He put his phone away and looked at the street.
The network was operational. The task architecture was different in kind, as the system had said it would be. Three nodes, three roles, one task distributed across all of them — his mechanism generating the structure, her mechanism identifying the social access point, Han Weiwei's research providing the destination that gave the whole operation a purpose.
The architect had designed this to be more efficient than three people working separately.
The architect was right.
He walked the two blocks to the subway entrance and went home. He had a Thursday morning session at seven and a Tuesday network task running through Mao Yingjie's contact, and the LP counter at 177,300 with 22,700 remaining to a threshold he had been building toward for months.
The city ran its evening.
He had a great deal to do.
Upstairs, in the National Bureau's fourteenth floor, Han Weiwei was still at her desk. He could feel her building's aggregate field from street level — not her specific signature at this distance, but the building's. She would be reviewing the session notes against her existing research, running the new data points against the pattern, finding where they fit.
She would find that they fit very precisely.
The probability field ran at its ambient Level 3 register.
Seven to ten days.
