Chapter 18 — The Bleed Model
The independent research room was small, well-lit, and located in the basement of the academy's magic theory building, which Raj found appropriate given that most genuinely interesting things in his experience happened in spaces that were not the main venue.
It had two measurement crystals on stands, a mana output tracking board along one wall, a worktable with enough space for two people to work without occupying each other's space, and a ward system baked into the walls and floor that contained any magical output within the room completely. Professor Maren had shown them both the safety protocols with the focused efficiency of someone who had supervised practical research before and had developed the protocols from experience rather than theory.
"Nothing above mid-level output without notifying me first," she had said, looking specifically at Raj when she said it.
"Understood," Raj said.
She had looked at him for one additional second that communicated clearly that understood and would comply were two different statements and she expected both. Then she had left them to it.
Sana set her notebook on the worktable and opened it to a section that had clearly been prepared in advance — pages of theoretical framework, diagrams of attribute interaction models, margin notes in three different colors of ink that corresponded to a coding system Raj had not been given the key to but could partially decode from context.
"The hypothesis," she said, uncapping her pen, "is that cross-attribute mana bleed in simultaneous dual-channel output follows a logarithmic interference pattern rather than the linear additive model in the textbook." She looked at him. "You have direct experience of this. I have the theoretical framework. Between us we should be able to build a practical model that is actually accurate."
"What do you need from me," Raj said.
"Run two attributes simultaneously at increasing output levels while I track the interference pattern on the measurement board." She gestured at the crystal stands. "Start low. Increase in controlled increments. I need clean data at each level before you move up."
Raj looked at the measurement board. It was a good setup — more precise than the placement crystal, designed specifically for research rather than assessment, capable of tracking subtle interactions between magical expressions rather than just registering presence and strength.
"Which two attributes," he said.
"Fire and wind first," she said. "Adjacent elements. Then we move to non-adjacent pairs." She paused. "Then three simultaneously."
"That will take more than one session," Raj said.
"I have booked the room every Tuesday and Thursday afternoon for the semester," she said, in the tone of someone who had considered this and prepared accordingly.
Raj sat down across from her, placed both hands on the table in a relaxed position, and began.
Fire and wind at low output was easy. He had been running these two together since the fourth month of training — they were the most natural pairing for his Red Mage configuration, fire for offense, wind for movement and detection. At low output the interference was minimal, a slight warmth where the two expressions overlapped at the channel junction, barely registering on the measurement board.
"There," Sana said, marking something. "First data point. Increase by one level."
He increased. The board registered the change. She marked it.
They worked through six levels methodically, Sana recording each data point with precise handwriting, occasionally asking him to hold a level steady while she took a second reading. She was a good research partner — clear instructions, no unnecessary conversation during active recording, questions that were specific rather than general. Raj found the rhythm of it comfortable in the way that well-structured work was comfortable.
At level seven the interference pattern changed.
Not dramatically. Not dangerously. A subtle shift in the quality of the overlap at the channel junction — less warmth, more vibration, a different texture that his trained sensitivity caught before the board registered it.
"Hold," he said.
Sana looked up immediately. "What happened."
"The interference pattern shifted at this level. It's not linear anymore." He kept the output steady, examining the sensation carefully. "It's — layered. The wind is pushing against the fire channel rather than running beside it. Like pressure against a wall rather than two streams in parallel."
Sana's pen was moving fast. "Can you quantify the pressure?"
"Not in numbers. In sensation — it is approximately the difference between resting your hand on a surface and pressing against it." He paused. "If I increase output from here without adjusting channel separation the bleed will be significant."
"How significant."
"Left sleeve level," he said.
She stopped writing. "I'm sorry?"
"I once set my left sleeve on fire from the inside at this output level without proper adjustment," he said. "It was an educational experience."
Sana looked at him for a moment with an expression that contained several things simultaneously — concern, fascination, the particular focus of a researcher who had just received a genuinely useful data point from an unexpected direction. She wrote — subject reports combustion incident at equivalent output level. Consistent with logarithmic model prediction.
"Adjust the channel separation," she said. "Whatever you did to fix it originally. I want to see the board reading before and after."
He adjusted. The vibration smoothed. The board reading shifted — a clean measurable difference that Sana marked with three exclamation points in the margin, which was the most expressive he had seen her handwriting get.
"There," she said quietly. "That is the logarithmic curve. The adjustment you made compensated for the non-linear pressure increase." She was looking at the board with the expression of someone watching a theoretical framework become a confirmed model. "Your instinctive adjustment is the practical solution to the problem the textbook doesn't acknowledge exists."
"Christine developed the formal version of that adjustment," Raj said. "I just — found it by accident and she formalized it."
"You found it," Sana said, looking at him with the direct clarity she applied to most things. "That matters. Theory follows practice more often than the textbooks suggest."
They moved to the next pair. Earth and lightning — non-adjacent, the most significant elemental contrast available. Raj brought both up carefully, mindful of the output level where the interference pattern had shifted on the previous pair.
The board immediately showed something different.
Not the gradual interference buildup of the fire-wind pair. A sharp clean division — the two attributes running in completely separate channels with almost no interaction at the junction point. Like two people walking in the same direction who had decided not to acknowledge each other.
"Interesting," Sana said. "The non-adjacent pair has less interference than the adjacent pair."
"Less natural sympathy," Raj said. "Fire and wind want to interact — they have related qualities. Earth and lightning have almost nothing in common so they ignore each other at the channel level."
"That completely inverts the textbook's prediction," she said.
"Yes," Raj said.
She wrote something that took up half a page.
The door opened at the forty minute mark and Kael walked in.
He stopped in the doorway taking in the scene — Raj at the table with both hands raised and two distinct attribute glows running simultaneously, Sana across from him writing at speed, the measurement board showing a pattern that was clearly not standard.
"The advanced class free period is in the courtyard," Kael said. "You two are down here."
"Research session," Sana said, without looking up.
"I can see that." Kael came in and looked at the measurement board with genuine interest, because whatever else Kael was he was someone who found magic interesting when it was doing something real. "What's the pattern?"
"Non-adjacent attribute interference," Raj said. "Earth and lightning. The interaction is significantly lower than adjacent pairs."
Kael looked at the board. Looked at Raj. Looked at the board again. "The textbook says non-adjacent pairs have higher interference because the elements have to work harder to coexist."
"The textbook is wrong," Sana said.
Kael pulled up the third chair and sat down with the complete naturalness of someone who had decided he was part of this now. "Show me the adjacent pair data."
Sana turned her notebook toward him without objection. Kael read it with more focus than he typically applied to scheduled theory classes, which said something about the difference between information in context and information in curriculum.
"The logarithmic curve," he said, pointing at the data. "At level seven the pressure shifts."
"Yes," Raj said.
"That's why high-output dual-attribute users blow their channels," Kael said. "They hit the curve and they don't know it's there so they don't adjust."
"Exactly," Sana said.
Kael leaned back. Looked at Raj with the expression he got when something had clicked into place. "You blew your channel once. Before you learned the adjustment."
"Left sleeve," Raj said.
"Not your channel?"
"I caught it early," Raj said. "External manifestation before internal damage."
Kael nodded slowly. "How early do you catch things."
It was a more specific question than it sounded. Raj considered it seriously. "Early enough," he said. "I have high sensitivity to internal mana state changes. I notice shifts before they become problems most of the time."
"Most of the time," Kael said.
"There have been exceptions," Raj said, thinking about a poison that had reached his elbow before Lily noticed, and not saying that part.
Kael looked at him for a moment with the particular expression he got when he was cataloguing something about Raj for later consideration. Then he looked at Sana. "You need a third data set. Someone with a confirmed dual attribute who isn't all-type. Different channel architecture, different interference pattern. Controls for Raj's unusual configuration."
Sana's pen stopped. She looked at Kael. "Do you know anyone with a confirmed dual attribute?"
"Me," Kael said.
Both of them looked at him.
"Fire primary, wind secondary," he said. "Low output on the wind — it's why my fine control is poor, the secondary channel is underdeveloped. Professor Maren said it in those words." He paused. "But it's there. Confirmed on the placement crystal." He looked at the measurement board. "You want clean data. You need a comparison case. Use me."
Sana looked at Raj. Raj looked at Sana.
"Professor Maren will need to know," Raj said.
"I'll tell her," Kael said, already rolling up his sleeve with the decisive energy of someone who had committed to a direction. "She'll say yes. She already thinks this research is the most interesting thing the advanced class has produced in three years, she told Tomis that, Tomis told me."
"Tomis told you that," Raj said.
"Tomis tells me everything," Kael said. "He apologizes for knowing things but then tells them anyway." He placed his hand on the measurement crystal with the confidence of someone who had made a decision and was now simply in the execution phase. "Ready when you are."
Sana looked at her notebook. At the board. At Kael's hand on the crystal. At Raj.
Something in her expression shifted — the pleased shift, the barely-contained one. She wrote unexpected third variable — fire-wind dual, single subject, non-all-type in the margin and underlined it twice.
"Fire first," she said. "Low output. On my mark."
They ran the session for another forty minutes with Kael as the comparison case and the data that came out of it was, in Raj's assessment, genuinely significant. Kael's dual-attribute interference pattern was entirely different from Raj's — higher baseline interaction, less clean channel separation, the logarithmic curve appearing earlier and more sharply. The difference between a natural all-type channel architecture and a primary-secondary configuration was measurable, specific, and completely absent from the textbook's theoretical model.
By the end Sana had filled twelve pages and was looking at her notebook with the expression of someone who had started the afternoon with a hypothesis and ended it with three new ones.
Kael was looking at the board with something that had moved past interest into the particular focused energy of someone who had found a problem that was personally relevant. "The curve appearing earlier in my configuration," he said. "That's why Professor Maren said my fine control is poor."
"Your secondary channel hits the logarithmic pressure point at a lower output than your primary expects," Raj said. "So when you try to use them together at your primary's natural output level the secondary is already in pressure and you lose precision."
Kael was quiet for a moment. "Can it be fixed."
Raj looked at Sana. Sana looked at her notes. "Theoretically," she said carefully. "If you developed the secondary channel independently until its output ceiling was closer to the primary's, the curve would shift. More balanced architecture — lower interference across the range."
"How long," Kael said.
"Months," Sana said. "Possibly a year. Consistent isolated secondary channel training."
Kael nodded once. The decisive nod of someone updating a training programme in real time. He stood and rolled his sleeve back down and looked at both of them. "Same time Thursday?"
"Same time Thursday," Sana confirmed.
He left. The room was quiet again.
Raj began releasing his attribute outputs cleanly, one by one, the shutdown sequence as automatic as the startup. The measurement board settled back to baseline. The ward walls stopped their faint resonance.
Sana was still writing. He waited until she reached a natural pause.
"Good session," he said.
She looked up. The pleased expression was fully present now, not contained at all — the expression of someone who had spent theoretical work confirming itself in practice and found the experience deeply satisfying. "Very good session," she said. "We have the core data for the first paper."
"First paper," Raj said.
"There will be more," she said simply. "This is only the adjacent and non-adjacent pairs. We have not touched triple-simultaneous output, dark attribute interactions, or the practical application framework for the adjustment technique." She paused. "Also Kael's secondary channel development is now a longitudinal study whether he intended it to be or not."
Raj looked at the twelve pages of notes. At the measurement board still showing the final data configuration from Kael's session. At the notebook with its three-color margin system and its two underlined unexpected variables.
"You were hoping for this," he said. "Not just the bleed model. The whole project. You came to the academy specifically for this kind of work."
Sana was quiet for a moment. Then — "My attribute is mid-level fire. Good control, limited ceiling. In a world where magical power is primarily attribute-based, my ceiling is set." She paused. "But theoretical research has no ceiling. And practical research that produces accurate models other mages can use—" she looked at her notes, "—that helps people regardless of what my personal output is."
Raj looked at her. He thought about a year of training where he had been the one with the lowest rank and had compensated through technique, tactics, precision, and the particular attention of someone who could not afford to be anything less than completely present. Different mechanism, same principle.
"Christine would like you," he said.
It came out without planning. He heard it land and watched Sana's expression register it — the specific register of hearing something that was clearly a real reference to a real person being offered as a real compliment.
She did not ask who Christine was.
"I will take that as a high recommendation," she said.
"It is," Raj said.
He gathered his things. At the door he paused. "Thursday," he said.
"Thursday," she confirmed, already writing again.
He walked out into the corridor and up toward the main building and thought about logarithmic curves and secondary channel development and twelve pages of notes and a girl who had decided that no ceiling on theory was a better deal than a ceiling on output.
He thought it was a very reasonable conclusion.
He pushed his glasses up his nose and went to dinner.
End of Chapter 18