The first exchange was clean.
Ye Shuangyu opened with the combination she'd been using for the past week — a feinted left strike that her tactical precognition told her he'd read as genuine, followed by the actual attack from an angle his redirection chains hadn't trained against. She ran the combination at seventy percent output. Training intensity, not combat.
He absorbed the feint's energy through the seed's standard intake. The kinetic force entered through his left forearm, routed through the seed's distribution pathways, and sat in the holding pattern he'd built over three weeks of daily sessions. Standard. The absorption efficiency meter in his peripheral awareness read sixty-three percent.
Sixty-three. Up from sixty-one two days ago.
The real attack arrived from the precognition-informed angle. He caught it — barely — and the absorption routing pulled the kinetic force into the same holding pattern. Two loads now, one genuine and one feinted, both available for redirection.
He redirected the combined load toward her left flank.
The redirection was deliberate. Conscious routing. He felt the fractional delay as the energy moved through the pathway he'd chosen rather than a pathway the seed chose for him.
She sidestepped the redirected force easily. "Still deliberate."
"Yes."
"Again."
---
The second exchange was faster. The third was faster still.
By the seventh exchange, the combat seed's routing had started doing something different.
He noticed it on the redirect. The energy from Ye Shuangyu's strike entered through his right shoulder — a high-angle attack she'd aimed at the absorption pathway he used least. The seed pulled the kinetic force through its distribution system. And instead of waiting for his conscious routing decision, it redirected.
Not to where he would have sent it. To where Ye Shuangyu was going to be.
He felt the difference immediately. The conscious routing delay vanished. The energy moved from intake to redirect in a single continuous flow, the pathway selection handled by the seed itself based on combat data it had been accumulating for three weeks.
The redirected force hit the space Ye Shuangyu's precognition had already told her to vacate. She moved — two seconds of foresight giving her the exit — but her expression changed.
"That was automatic," she said.
"Yes."
She reset her stance. Her eyes had the focus she wore when the precognition was running at full attention. "Do it again. I want to see if the routing is consistent."
The eighth exchange. Her attack came from below — an upward strike aimed at the absorption pathway along his ribs. The seed caught it, routed it, and redirected without his conscious input. The energy moved toward the predicted position with the same automatic flow.
"Consistent," she said. "The seed is reading my movement patterns and predicting redirect targets." She paused. "It's using the training data."
"Three weeks of daily combat assessment against the same partner," he said. "It built a model."
"That model will need to generalize," she said. "You can redirect against me because the seed has my patterns. Against an unfamiliar opponent, it'll revert to conscious routing until it builds a new model."
"How long to build a model against a new partner?"
"Based on the timeline against me — two to three sessions for the basic pattern recognition. Longer for the automatic redirect threshold."
He filed this. The combat seed's automatic routing was partner-specific. Useful against known opponents, limited against unknown ones. The generalization problem would need systematic work.
"Multi-partner session," she said. "Chen Wei and Tan Xueying are in the rotation today."
---
Chen Wei arrived first. B-rank physical enhancement, the kind of straightforward combat profile that made him good sparring material — consistent output, predictable energy signatures, the reliable baseline that training assessments needed.
The exchange against Chen Wei started with conscious routing. The seed's automatic system didn't have Chen Wei's patterns. He felt the delay return — the fractional pause between absorption and redirection where he had to choose the pathway manually.
But the delay was shorter. Three weeks of training had made the conscious routing faster even when the automatic system wasn't engaged. The absorption efficiency against Chen Wei's strikes read sixty-five percent.
He redirected three of five absorbed loads successfully. Not automatic. Faster than manual.
"Your intake angle on the left side has improved," Chen Wei said. "Two weeks ago you were losing twenty percent of the absorbed energy on left-side strikes."
"The pathway developed," he said.
"It's developing fast," Chen Wei said, and there was something in his voice — the specific register of a B-rank combat specialist watching someone at his own tier pick up combat mechanics at a rate that wasn't standard.
Tan Xueying joined the rotation. Her combat type was different — energy projection rather than physical enhancement, which meant the kinetic force the seed absorbed had a different quality. Energy projection strikes carried a frequency component that physical enhancement didn't.
The first exchange against Tan Xueying's projected energy: the seed absorbed it and the efficiency dropped. Fifty-four percent. The frequency component was harder to route.
He adjusted. Second exchange: fifty-seven. Third: fifty-nine.
"The absorption adapts to different energy types," Ye Shuangyu said from the observation position. "But the adaptation curve is slower for energy projection than physical enhancement."
"The frequency component," he said. "The routing pathways were built on kinetic force. Energy projection has a waveform that doesn't map to the same pathways."
"So you need pathway development for each energy type," she said.
"Or a generalized pathway that handles multiple input types," he said. "Which would be less efficient but more versatile."
"The generalization problem again," she said. "Build it systematically. Run against Tan Xueying's energy projection for the next two sessions, then alternate."
He nodded.
The multi-partner session continued. Ye Shuangyu called the rotation — Chen Wei, Tan Xueying, back to Chen Wei, then both simultaneously. The simultaneous exchange was new. He'd absorbed from two sources before, but the seed had to route two different energy types through the same distribution system at the same time.
He managed it for four seconds.
On the fifth second, the absorption threshold hit something.
The kinetic force from Chen Wei's physical strike and the energy projection from Tan Xueying's waveform strike entered the seed's system simultaneously. The combined load crossed a threshold he hadn't reached in single-partner training. The seed's distribution pathways routed both inputs — and the combined energy density in the holding pattern interacted with the probability field.
He felt it as a vibration. Not physical. The Luck Aura's ambient assessment — the constant background calculation of probabilities in his environment — stuttered. For approximately half a second, the probability field's output dropped to zero and then spiked to a level he'd never registered before hitting zero again.
The combat seed's holding pattern dumped its stored energy in an uncontrolled discharge. The kinetic force went sideways — not toward any target, just out. Chen Wei stumbled back. Tan Xueying's next projected strike passed through empty air where he'd been standing before the discharge moved him two meters to the right.
The training room went quiet.
"Stop," Ye Shuangyu said.
He was standing two meters from where he'd been, the probability field's ambient calculation restarting like a system that had rebooted. The Luck Aura's warm texture was back, but there was a lag — the assessment running at reduced resolution for a few seconds before returning to normal.
"What was that?" Chen Wei said.
"Resonance interaction," Ye Shuangyu said. She crossed the room to where Chen was standing. The tactical precognition was running — he could tell from her eye movement, the specific tracking pattern she used when she was reading a situation two seconds ahead and comparing it to what was happening in real time. "The combat seed's stored energy at dual-input threshold density interacted with his probability field."
"My field dropped to zero," he said. "Then spiked, then dropped again."
She looked at him. "How long was the zero state?"
"Half a second. Maybe less."
"Half a second without the probability field's ambient assessment," she said. "In a real combat situation, that's a window where luck isn't running."
He understood the problem immediately. The Luck Aura was always active. Level 4 ambient probability assessment, running constantly, adjusting outcomes in his favor by degrees too small to notice individually but significant in aggregate. If the combat seed's energy storage at high density could disrupt that field — even for half a second — it created a gap in the probability coverage.
Half a second where he was operating without the system's passive assistance.
In a B-rank fight, half a second was survivable. Against an A-rank opponent, it might not be. Against an S-rank opponent, it was a death sentence.
"The combat seed and the probability field are both system-purchased," she said. "They're running on the same host architecture. At low energy density, they don't interact. At the dual-input threshold density, the seed's stored kinetic energy creates a resonance frequency that matches the probability field's operating frequency."
"How do you know it's a frequency match?"
"Because the field didn't just drop — it spiked first," she said. "The resonance fed energy into the probability field before the field's safety systems shut it down to prevent overload. Spike, then zero, then restart."
He looked at the training room floor where the uncontrolled discharge had left a scuff mark. The energy hadn't been directed. It had just gone.
```
[COMBAT SEED INTERACTION — DIAGNOSTIC]
[DUAL-INPUT THRESHOLD: EXCEEDED AT COMBINED DENSITY OF 2,340 UNITS]
[PROBABILITY FIELD RESONANCE: DETECTED — FREQUENCY OVERLAP AT 0.73 HARMONIC]
[FIELD DISRUPTION DURATION: 0.4 SECONDS]
[NOTE: THE RESONANCE IS A CALIBRATION ISSUE. THE COMBAT SEED AND THE PROBABILITY FIELD WERE PURCHASED SEPARATELY AND INSTALLED WITHOUT CROSS-SYSTEM CALIBRATION. AT LOW ENERGY DENSITIES, THE SYSTEMS OPERATE INDEPENDENTLY. AT HIGHER DENSITIES, THEY SHARE HARMONIC FREQUENCIES THAT CREATE INTERFERENCE.]
[NOTE: THE CALIBRATION CEILING IS CURRENTLY AT 2,340 COMBINED INPUT UNITS. EXCEEDING THIS DENSITY WITHOUT CALIBRATION ADJUSTMENT WILL TRIGGER THE RESONANCE EVERY TIME.]
[NOTE: THIS CEILING LIMITS YOUR EFFECTIVE DUAL-INPUT ABSORPTION TO APPROXIMATELY 67% OF THE COMBAT SEED'S THEORETICAL MAXIMUM. THE REMAINING 33% IS INACCESSIBLE UNTIL THE HARMONIC FREQUENCY OVERLAP IS RESOLVED.]
[NOTE: RESOLUTION REQUIRES SYSTEMATIC CROSS-SYSTEM CALIBRATION TRAINING. ESTIMATED TIMELINE: 4-6 WEEKS OF TARGETED SESSIONS.]
```
He read the diagnostic. Four to six weeks. The calibration ceiling at sixty-seven percent of theoretical maximum. A third of the combat seed's capacity locked behind a resonance problem he hadn't known existed until five minutes ago.
The side task appeared.
```
[SIDE TASK: THE RESONANCE INTERACTION REVEALED THE 0.73 HARMONIC OVERLAP. YE SHUANGYU'S TACTICAL PRECOGNITION CAN MAP THE EXACT ENERGY DENSITY THRESHOLDS WHERE THE RESONANCE TRIGGERS. RUNNING A SYSTEMATIC THRESHOLD-MAPPING SESSION TODAY WOULD ESTABLISH THE CALIBRATION BASELINE. REWARD: 4,400 LP + [COMBAT CALIBRATION: BASELINE ESTABLISHED]]
```
"The threshold density," he said to Ye Shuangyu. "Can you map it?"
She'd already been running the assessment. "The precognition doesn't read energy density directly. But I can read the combat outcome probability shifts that correspond to the resonance trigger. When your probability field drops, the precognition registers it as a sudden change in your combat outcome distribution."
"You can see when I lose the luck."
"I can see when the combat math changes around you," she said. "Yes."
He looked at Chen Wei and Tan Xueying. Both were waiting. Chen Wei had the expression of someone who'd seen something he didn't fully understand but knew was significant. Tan Xueying was watching Ye Shuangyu, following the analysis.
"Threshold mapping session," Chen said. "Graduated energy density. I absorb increasing combined loads until the resonance triggers, and you map the exact trigger points."
"It'll be uncomfortable," she said. "The resonance discharge at each trigger point will dump your stored energy uncontrolled."
"Controlled enough to stay in the training room?"
"At graduated increases, yes. The discharge intensity scales with stored energy. We keep the loads small."
He turned to Chen Wei and Tan Xueying. "Graduated dual-input. Start at fifty percent of what you were running before. Increase by five percent per exchange."
Chen Wei nodded. Tan Xueying adjusted her output accordingly.
The threshold mapping took forty minutes. Eight trigger points identified between 2,340 and 3,100 combined input units. Each one caused the probability field disruption — the spike-zero-restart sequence — and a small uncontrolled discharge. Each discharge was smaller than the first because the stored energy loads were graduated.
Ye Shuangyu mapped each trigger point against her precognition data. The combat outcome probability shifts she registered at each resonance event. By the end, she had a baseline curve.
"The harmonic overlap isn't uniform," she said. "It triggers at specific density bands, not at a single threshold. There are gaps between the bands where you can operate at higher density without triggering the resonance."
"Operating in the gaps," he said.
"Until the cross-system calibration resolves the overlap," she said. "Yes. You have safe density bands where you can absorb at higher rates, and danger bands where the resonance will trigger. Learn the bands."
`[SIDE TASK: COMPLETE. +4,400 LP. COMBAT CALIBRATION BASELINE ESTABLISHED]`
He looked at the mapping data Ye Shuangyu had recorded. The density bands. The gaps. Four to six weeks of calibration work to resolve the harmonic overlap permanently.
"The sixty-seven percent ceiling," he said. "Can I push past it using the safe bands?"
"Theoretically, yes," she said. "The safe bands allow you to exceed the trigger threshold by routing through non-resonant density ranges. But the routing has to be precise. You'd be managing your absorption energy density in real time during combat."
"Another conscious routing problem."
"A harder one," she said. "The partner-specific automatic routing was pattern recognition. This is energy density management at the physics level. Different skill entirely."
He sat on the training room bench. His arms ached from the repeated resonance discharges — a physical aftereffect of having his stored kinetic energy dumped uncontrolled eight times in forty minutes.
Two limitations exposed in one session. The partner-specific routing needed generalization. The harmonic resonance needed calibration.
"The calibration training," he said. "We add it to the rotation."
"Wednesday and Friday," she said. "The threshold mapping gives us a structure. Graduated density work in the safe bands, with resonance tolerance development in the danger bands at low intensity."
"How long before I can fight at dual-input threshold without the disruption?"
She looked at the mapping data. Ran the precognition against some internal calculation he wasn't part of.
"Six weeks for basic resonance management," she said. "Longer for combat-speed calibration. You won't have this solved before the quarterly review."
He filed this. The quarterly review — six weeks, the Lin Family negotiation reassignment opportunity. He wouldn't have the combat calibration issue resolved by then. Wouldn't need to — the quarterly review was institutional, not combat.
But the gap was noted. The S-rank gap between him and Lin Zhengyue was already the primary limitation on MQ3's timeline. A combat ceiling at sixty-seven percent of theoretical maximum made that gap wider.
"The systems interact," he said. "The combat seed and the probability field. They weren't designed to run together."
"They were designed to run independently," Ye Shuangyu said. "The architect built the probability field. The combat seed was a shop purchase. Two separate engineering philosophies installed on the same host."
"Integration problems," he said.
"Integration opportunities," she said. "The resonance spike — before the field shut down — pushed your probability field output briefly above anything I've seen from Level 4. The harmonic frequency doesn't just interfere. At the right density, it amplifies."
He looked at her.
"The spike," he said.
"The spike," she said. "For approximately one tenth of a second before the safety cutoff, your probability field was running at what I'd estimate as Level 5 output intensity."
She let this sit.
Level 5 required one million LP total earned. He was at 481,500. The spike had pushed the probability field to Level 5 output for a tenth of a second through resonance amplification.
Not sustainable. Not controllable. Not safe. But the interaction existed.
"The calibration work," she said. "It's not just about preventing the disruption. If you can learn to manage the resonance rather than suppress it—"
"The amplification becomes a tool," he said.
She didn't confirm. She didn't need to. The precognition had already shown her the possibility she was describing, two seconds ahead of the conversation, and she'd chosen to share it.
The training room's afternoon light moved across the floor. His arms still ached. The mapping data sat on the bench beside him, eight trigger points and their corresponding density bands, a calibration problem that would take weeks to solve.
He'd come in at sixty-one percent absorption efficiency and discovered a ceiling he hadn't known existed.
The complication was real. So was what lived on the other side of it.
