Sleep wouldn't come.

  Sylas lay in his damp bed — the roof still leaked, the maintenance request still somewhere in the Academy's filing system quietly not being acted upon — and stared at the water-stained ceiling while his mind did what it always did when left unattended with a problem it hadn't finished solving. It kept working. It turned the problem over, examined the parts, ran scenarios, returned to the parts that hadn't been fully resolved and prodded them until they gave up something useful or confirmed that they weren't going to.

  Three days until the examination. Three days until everything he had spent the past month carefully constructing would be evaluated in earnest by people whose specific job was to evaluate it.

  The practice test had confirmed the written component strategy worked. Sixty-four percent, exactly calibrated, accepted by Graves as adequate progress without comment. But the practice test had been only the written section. The real examination had three parts: written theory, practical cultivation demonstration, and mana projection. Different skills, different observation conditions, different failure modes, different ways for the performance to break down.

  He needed to pass all three. He needed to pass all three with precisely calibrated mediocrity that told a consistent story about a student who was working at the edge of his ability and just barely making it. Three sections, each observed by different instructors, each measured by different criteria, each requiring a performance that was internally coherent with the others.

  After lying still for what felt like a substantial portion of the night, Sylas accepted that sleep was not the productive use of the next few hours and sat up carefully, trying not to disturb Arthur, whose breathing had the settled rhythm of someone properly asleep.

  He reached for his satchel. Paper, charcoal, the cultivation manual for reference. Moved to the desk with the careful quiet of someone navigating a shared space in the dark. Lit the small oil lamp they kept for studying, turning the wick down low enough to read by without flooding the room with light.

  Time to do what he did best: plan, calculate, optimize.

  He started with the fundamental constraint and wrote it at the top of the first page, because writing things down made them concrete in a way that was useful for the kind of thinking he was about to do:

  TARGET: 30–35% OVERALL

  Thirty percent was the passing threshold. Below it and you were marked for Reclamation, the machinery beginning to process you before the examination had properly ended. Above thirty-five — comfortably above it, demonstrably and consistently above it — and you became a student on an upward trajectory, which meant increased scrutiny, increased expectations, increased probability of someone noticing that your performance didn't match your history.

  Thirty to thirty-five. Pass without distinguishing yourself. The corridor he needed to stay inside.

  Now the breakdown. He wrote the section weightings from memory, cross-referencing against Arthur's notes from previous examinations to confirm:

  Written Theory: 20% of total score. Practical Cultivation: 50% of total score. Mana Projection: 30% of total score.

  The practical cultivation section was worth half the examination. That was the load-bearing component — the one where failure was most dangerous and the one where his actual capability was most radically divergent from his intended performance. Five minutes of continuous mana circulation, observed directly by instructors who had Mana Sight active. Fifty percent of the score riding on five minutes of performing inadequacy while actually being competent.

  He started running scenarios on the second half of the page, working methodically through combinations.

  SCENARIO 1: Written 65% = 13 points. Practical 40% = 20 points. Projection 0% = 0 points. Total: 33%. The mathematics worked. The zero on projection did not. Completely failing a section looked like freezing under pressure or giving up, both of which invited the wrong kind of attention — not the scrutiny of unusual competence, but the scrutiny of psychological instability or deliberate non-participation, both of which would result in a conversation he didn't want to have.

  SCENARIO 2: Written 50% = 10 points. Practical 40% = 20 points. Projection 10% = 3 points. Total: 33%. Better distribution. The lower written score was achievable — he'd just need to miss more questions than on the practice test, including some easier ones, which risked looking inconsistent with the practice result. He noted the risk.

  SCENARIO 3: Written 60% = 12 points. Practical 35% = 17.5 points. Projection 15% = 4.5 points. Total: 34%. This felt right in the way that a properly designed system felt right — balanced, internally consistent, each component telling the same story as the others. Moderate performance across all sections, nothing exceptional in any direction, no dramatic failures that would require explanation. A student working hard and achieving adequately.

  He circled Scenario 3 and began the detailed breakdown.

  WRITTEN THEORY — Target: 60%

  Twenty-five questions, based on the practice test format. Sixty percent meant fifteen correct, ten wrong. The strategy was already established and tested: answer all the foundational questions correctly to demonstrate basic engagement with the material, miss the complex integrative questions that required holding multiple concepts simultaneously or understanding the theoretical framework's internal logic. Pattern of answers should read as a student who had covered the basics but hadn't yet internalized the deeper structure.

  He could do that. He had done that. The variation from the practice test's sixty-four percent to the target sixty percent was small enough to be within normal day-to-day variation rather than a detectable decline.

  He moved to the harder problem.

  PRACTICAL CULTIVATION — Target: 35%

  The practical was graded on four factors, each requiring its own strategy:

  Completion: whether the student could maintain circulation for the full five minutes without stopping. This was binary — either you completed it or you didn't. Partial completion scored poorly and also told a story he didn't want to tell. He would complete the five minutes. Exactly five minutes, not a breath longer, the way a student who had pushed to his limit would stop the moment the minimum was met.

  Stability: whether the circulation held steady or showed fluctuation. The examiners would be watching for losses of control — moments where the mana circulation broke and had to be restarted, moments where it degraded in quality, moments where the student clearly struggled to maintain what they had achieved. A thirty-five percent score required showing instability. He planned two controlled instability events: a brief circulation disruption around the two-minute mark as though his focus had slipped, and a second one around the four-minute mark as though the sustained effort was reaching his limit. Both would be recoveries rather than full collapses — recoveries demonstrated effort and persistence, which scored better than giving up.

  Efficiency: the ratio of gathered mana to stored mana, measured by how much reached the dantian versus how much was lost in transit. His optimized technique produced roughly twenty percent loss. The standard technique produced thirty to forty percent loss. A struggling student using the standard technique inconsistently might produce fifty to seventy percent loss. He needed to target sixty percent loss — three times his actual technique's waste rate, achieved by deliberately routing a portion of his gathered mana through unnecessary pathways where it would dissipate before contributing to storage.

  He thought about the mechanics of this. He could route mana to his extremities on the first circulation loop and let it disperse in the capillary channels — the same pathway Mae had taught the study group to pursue, the tingling sensation that signalled energy loss. He could also create controlled turbulence at the Solar Plexus Node by deliberately introducing a phase mismatch in the energy he channelled through it, which would cause localised dissipation without disrupting the overall circuit. Both methods would produce the right level of visible inefficiency without requiring him to actually break his circulation.

  If you find this story on Amazon, be aware that it has been stolen. Please report the infringement.

  Form: whether the student appeared to be following the standard technique. This was the observation he was most confident about — he had practised the performance of the standard technique's seventeen-point sequence for two weeks, and the examiners were reading timing and visible effort rather than the specific pathway the mana was taking. Artificial pauses at the standard checkpoints, appropriate facial expression and breathing, the visible physical strain that Sylas had catalogued at length. He had this component.

  He wrote "COMPLETED" next to Practical Cultivation and moved to the third section.

  MANA PROJECTION — Target: 15%

  External mana manipulation — projecting energy beyond the body's meridian system and directing it at a target. This was the section he knew least about in terms of examination format, because Arthur's notes were thinner here. What he knew: students were asked to project mana at a fixed target and the result was graded on distance, force, and control.

  Fifteen percent was the minimum meaningful score — just enough to demonstrate that he had some external projection capability without demonstrating anything interesting about its quality. He calculated the parameters:

  Distance: one to two metres. Short enough to look like a student who hadn't developed strong projection range, long enough to confirm the capability existed at all. Zero distance — projecting at something immediately in front of him — would look like he was avoiding the test.

  Force: enough to visibly move the target. The target would almost certainly be a lightweight marker of some kind — the Academy used standardised examination equipment. He needed to move it, not propel it dramatically across the room. A marker that shifted position rather than flew was the right result.

  Control: deliberately scattered. Not random — random scatter had a specific visual texture that was different from the scatter produced by imprecise control over a real but underdeveloped ability. He needed to project with the quality of someone who had the capability but not the refinement, not the quality of someone who had never tried. The distinction was in how the energy dispersed: early scatter from poor channel alignment, not late scatter from insufficient force.

  He could produce all of this. He spent a few minutes sketching projection diagrams on the third page — the clean, focused projection pattern he was capable of, and the degraded version he would perform, with notes on how to degrade each element specifically.

  The lamp's flame bent slightly as a draught moved through the room from the gap under the door. Sylas looked at what he'd produced: three pages of calculations, diagrams, contingencies, specific performance targets for each graded component of a three-part examination. Exact numbers. Backup numbers for if conditions varied. Notes on the mechanics of performing each failure mode convincingly.

  He had put more analytical work into planning how to score thirty-four percent than most students in this institution put into planning how to score their best. The irony had a specific shape that he was not fully sure how to feel about.

  He turned to a fourth page and ran contingency scenarios, because thoroughness was the thing he defaulted to when he was anxious and he was anxious, and thoroughness was at least productive anxiety.

  If the written section was harder than the practice test — more complex questions, higher proportion of integrative problems — he might naturally score lower than sixty percent without any deliberate miscalibration, which would require him to answer more questions correctly than intended to compensate. He noted the adjustment: if written feels harder, answer two additional complex questions correctly to maintain the sixty percent target.

  If the practical grading was more lenient than expected — if the examiners were less exacting about efficiency or stability — his thirty-five percent target might produce a higher score than intended. He noted the adjustment: if practical examiners seem lenient, increase the efficiency loss to seventy-five percent and add a third controlled instability event.

  If the mana projection section was weighted differently than Arthur's notes suggested — he had been going on secondhand information, and examination formats shifted — he might need to adjust his projection performance. He noted: if projection appears more heavily weighted, add a second projection attempt at marginally longer distance.

  "Sylas?"

  He turned. Arthur was sitting up in bed with the specific expression of someone woken by ambient light who was trying to assess whether they needed to be fully awake or could return to sleep. The lamp's glow had apparently been sufficient to pull him up from wherever he'd been.

  "Sorry," Sylas said, immediately. "Didn't mean to wake you."

  "What time is it?" Arthur squinted at the window, which offered nothing useful — the darkness outside was the uniform darkness of middle-of-night rather than anything that helped with time estimation. "It's very dark. What are you doing?"

  "Working through the exam," Sylas said. "The section breakdowns. Couldn't sleep."

  Arthur got out of bed with the slow coordination of someone who has decided that going back to sleep without addressing the thing he's seen would be difficult. He came over to the desk and looked at the pages spread across it. The calculations. The diagrams. The percentage targets with their contingency adjustments.

  He was quiet for a moment, reading it.

  "Gods, Sylas." His voice had the quality of someone who was both impressed and slightly alarmed. "This is — you've mapped out every section. Weightings, targets, contingencies. You've calculated exactly what you need from each component to reach the overall score." He looked up. "This is the most thorough exam preparation I've ever seen from anyone."

  Sylas looked at his pages of carefully engineered failure. At the hours of work he had just invested in determining the precise mechanics of scoring thirty-four percent.

  "Just trying not to die," he said, which was the most accurate summary available.

  Arthur took this as anxiety, which was the only reasonable interpretation from where he was standing. "You're not going to die. Look at this — anyone who puts this much thought into their preparation is going to pass. You've covered every angle." He put a hand on Sylas's shoulder with the uncomplicated warmth of someone who cared about another person and wanted them to know it. "But you should sleep. You can't perform well if you've been up all night."

  "You're right," Sylas said. "Give me a few minutes to finish this thought."

  Arthur returned to his bed, and was asleep again quickly — he had the enviable quality of someone who could re-enter sleep without extended negotiation. Sylas looked at the pages one more time and ran through the key numbers, committing them to memory with the care of someone who might not have access to these notes at the moment they needed them.

  Thirty-four percent overall. Sixty percent on written — fifteen correct, ten wrong, failures clustered in the integrative questions. Five minutes exactly on practical, two controlled instability events at the two-minute and four-minute marks, sixty percent mana waste through deliberate extremity routing and Solar Plexus turbulence, standard-technique form performance throughout. One-to-two metre projection on the mana section, minimal force, early scatter.

  He went over it once more. Then again. The numbers sat solidly in his memory in the way that numbers did when they had been arrived at through a process thorough enough to make them feel inevitable rather than chosen — not memorised so much as confirmed, each one the end result of an argument that had been made and accepted. He had done this kind of night work before, in a previous life, in the hours after a particularly bad quarterly review or before a presentation that mattered, sitting at a desk while the building was quiet and running numbers until the anxiety had somewhere specific to put itself. The numbers never fully solved the anxiety. But they gave it a shape, and a shaped problem was a different kind of problem than a formless one.

  He tucked the papers into his satchel carefully. He would need to dispose of them before the examination — finding pages of detailed strategic-failure calculations in a student's possessions would produce questions for which he had no useful answers and which would, at minimum, require an explanation that coherently accounted for how a student in Foundation Establishment had developed a working model of the examination's scoring architecture. He had not developed such an explanation and did not intend to need one. The pages would go into the Academy's waste system tomorrow evening. He extinguished the lamp and climbed back into the damp bed.

  The ceiling dripped once, faithfully.

  Three days. He had the plan. He had the numbers. He had the performance rehearsed to the level of near-automaticity. He had, in the most literal sense possible, done his homework.

  What remained was executing it under observation, which was the part that could not be fully rehearsed in a room alone at night. Five minutes of deliberate inefficiency in front of people trained to evaluate cultivation, who had Mana Sight active and who were specifically looking for the things he needed to appear to be doing. One page of carefully miscalibrated answers, produced in an examination hall with other students around him and proctors watching for behaviour that did not look like genuine effort. One short sloppy mana projection at a target a metre away, in a room where the quality of projection attempts was the subject of active professional attention. The performance would need to hold under conditions that the rehearsal had approximated but could not replicate exactly.

  The World's Most Adequate Cultivator, planning his adequate success with the precision of an engineer — except that what he was engineering was not a structure designed to stand but a structure designed to hold together at exactly one point above the minimum, indefinitely, without anyone noticing that it could have been built stronger.

  He lay in the dark and listened to Arthur breathe and to the ceiling drip and to the sounds of the Academy settling around them, the old building doing what old buildings did in the quiet hours.

  Three days. He could do three days.

  He closed his eyes and, eventually, slept.