The stylus I'd made with Garret was wearing out.

  Bronze alloy doesn't fail suddenly; it erodes. The tip had broadened fractionally over months of inscription work, the precise point I'd shaped for fine channel carving losing its geometry to friction and mana heat and the simple physics of a hard tool pressed repeatedly into harder stone. The broadening was subtle enough that I hadn't noticed it happening. What I'd noticed was the results getting worse: channels that should have been clean-edged were developing micro-fissures at the margins, and the mana flow through those channels was turbulent rather than smooth. I'd spent two days troubleshooting my technique before I caught it. Adjusted my grip. Ran the same test channel six times and got the same ragged edge six times. The technique was fine. The tool wasn't.

  Two days. Two days of assuming the problem was me rather than the instrument, which told me something unflattering about my default mode of self-criticism, and something useful about the way tool degradation hides inside gradual decline. In my old life I'd worked with people who spent months optimizing their process before someone thought to check the equipment. The principle was identical: you trust the tool and doubt the hand, because the hand is the thing you can control, and controlling what you can control feels like progress even when the actual problem is sitting right there in your grip, wearing down one micron at a time.

  I needed a replacement. More precisely, I needed a better original. The bronze alloy Garret and I had settled on was a compromise between hardness and workability; it held a point well enough for basic inscription, but my work wasn't basic anymore. The resonance experiments, the channel geometry studies, the formation observation work I'd been doing on Bren's maintenance rounds, all of it demanded finer control than a hand-forged bronze stylus could sustain. I'd watched Bren work during those rounds, and his inscription tools were different: harder alloy, finer gauge, the tips ground to a precision that my workshop version couldn't match. Professional grade. The kind of thing that should exist somewhere in the castle's supply system.

  So I went looking for the supply system. Which was how I discovered the administrative wing, and the administrative wing discovered me.

  I'd been vaguely aware that the castle had an administrative apparatus. You couldn't live in a building this size without noticing the clerks in the corridors and document carts being wheeled between offices, the particular type of person who walked with the brisk purposefulness of someone for whom the building was not a residence but a workplace. Servants moved differently from administrators. Servants flowed around corners and through doorways like water finding its level; administrators walked in straight lines, carrying things, heading somewhere specific. I'd noticed the difference years ago and filed it under background observation.

  The castle's administrative offices occupied a corridor on the second floor of the central building, between the household quarters and the formal reception rooms. I'd passed through this corridor hundreds of times without paying particular attention to the doors that lined it. They were the infrastructure of governance: the offices where supplies were tracked, budgets managed, schedules coordinated, and the thousands of daily decisions that kept a castle of this size operational were processed, documented, filed, and occasionally acted upon. The corridor smelled like paper and lamp oil and the specific dry warmth of rooms where someone kept a fire going year-round to prevent ledger pages from absorbing moisture. In my old life, government buildings had smelled like floor wax and recycled air. Different materials, same underlying function: the scent of a place where decisions were made slowly and recorded permanently.

  I found the correct office by the method of asking a servant, who directed me to another servant, who pointed at a door labeled "Clerk, Educational and Craft Materials" in lettering that had been professionally painted at some point in the past and had since been subjected to the administrative equivalent of erosion: partially covered by a posted notice, slightly faded, and flanked by two additional signs that appeared to modify its meaning in ways I couldn't parse from the corridor. One sign referenced a revision to form classifications dated fourteen years ago. The other appeared to be a correction to the first sign.

  The door was open. The office behind it was small and organized with the density of a space that had been accumulating function for longer than it had been accumulating furniture. Shelves filled with ledgers, arranged not by any system I could identify but with the internal logic of someone who knew exactly where everything was and saw no reason to make that knowledge legible to visitors. A desk covered in forms. Behind the desk, a man of indeterminate middle age whose bearing suggested decades of processing requests that were, to him, interchangeable in their particulars. He had the posture of a person who had found the exact position in which his chair, his desk, and his reach to the surrounding drawers were all optimized, and who had not adjusted that position in years.

  "Good morning," I said. "I'd like to request an inscription stylus."

  The man looked up. Took in my age, my clothing (which identified me as a member of the royal household, though not which member), and the nature of the request. None of this appeared to interest him beyond the data it contributed to form selection.

  "Educational or craft?" he said.

  "Inscription stylus," I repeated. "Fine-gauge, hard alloy tip. For carving inscription channels in stone and metal substrates."

  "Yes, but is the request educational or craft?" He had a form in front of him already, pulled from a drawer with the practiced efficiency of someone who had been doing this longer than I'd been alive. In either of my lives. "Educational materials route through one process. Craft supplies through another."

  I considered the question. The stylus was for my craft practice, which was part of my education, which I conducted in workshops, which were managed under a third administrative category. In a sane system, these categories would overlap. But I was standing in an administrative office in a castle that had been running for four hundred years, and the categories had been defined by people who saw no reason for overlap, and who were possibly still alive to defend that decision.

  "It's for inscription work that I do as part of my independent study programme," I said.

  "Independent study." He wrote something. "That falls under Educational Materials, subcategory Supplemental Tools. The request routes from this office to Adjunct Marun for categorization review, then to Educational Oversight under Lord Senrel's division for approval."

  "How long does the approval take?"

  "Standard review cycle is six weeks."

  I stood there for a moment. Six weeks. In my old life, I'd once waited fourteen months for a building permit, so the number itself wasn't shocking. What was shocking was the ratio. Six weeks for a stylus. A stylus that weighed less than a spoon and cost less than a decent meal, routed through three offices and a formal review cycle that could have processed a request for a new wing of the castle at roughly the same speed.

  "Six weeks," I said. "For a stylus."

  "For any non-standard educational materials request." He was writing on the form with the focused attention of someone whose job was the form, not the request. "Standard items are pre-approved on the quarterly supply list. Non-standard items require individual review."

  "An inscription stylus isn't on the standard list?"

  "Inscription tools are categorized as specialty craft equipment. They aren't part of the standard educational curriculum."

  "Because inscription isn't taught as part of the standard curriculum."

  "Correct."

  I could feel the shape of the logic. It wasn't wrong, exactly. Inscription tools weren't standard educational equipment because inscription wasn't a standard subject. They weren't pre-approved because nobody had added them to the approval list, because nobody teaching the standard curriculum needed them. The system was functioning as designed; the design simply hadn't anticipated a twelve-year-old prince who needed professional-grade inscription tools for self-directed formation research. Why would it have? How often does that happen? Probably never. The form didn't have a checkbox for it, and in an administrative system of this age, if there's no checkbox, the thing doesn't exist.

  "What information do you need from me?" I asked.

  "Name, household designation, educational supervisor's acknowledgment, description of the item requested, justification for non-standard classification, and signature of a sponsoring household member of sufficient rank."

  "What rank is sufficient?"

  "Third tier or above. A senior household official, a prince of the fourth rank or higher, or a designated educational authority."

  Marvenn. I'd need Marvenn's signature for a stylus.

  "Thank you, Clerk..."

  "Orveth," he said.

  "Clerk Orveth. I'll return with the completed form."

  He nodded and went back to his ledger. The interaction had cost him perhaps three minutes of attention; I could see the entry he'd made already settling into the administrative record like sediment on a riverbed. One more request. One more form. One more six-week cycle beginning its slow rotation through the machinery.

  I came back three days later with the completed form. Marvenn had signed it, though the signing had involved a conversation I hadn't entirely anticipated.

  I'd found him in his study, which was where Marvenn could usually be found in the late afternoon: a room organized with the careful neutrality of someone who didn't want his workspace to reveal anything about his priorities. Books shelved by category, correspondence in neat stacks, nothing personal on the desk except a teacup that was always present and always half-full. He read the form through once, completely, before looking up.

  "Six-week review cycle," he said.

  "For a stylus, yes."

  His eyebrow did something. Not quite a raise; more of a lateral shift, as if the eyebrow itself were reviewing the information and finding it procedurally correct but operationally absurd. Marvenn's eyebrows were, in my experience, more honest than his words. His words were always careful. His eyebrows occasionally editorialized.

  "That's the standard process," he said.

  "You don't think it's excessive?"

  "I think it's the process. The process exists for reasons." A pause. He set the form down on his desk and lined it up with the edge, which was the kind of thing Marvenn did when he was thinking and didn't want his hands to be idle. "Some of those reasons are still relevant."

  "And the others?"

  He gave me one of his looks. I'd been cataloging Marvenn's looks since I was old enough to notice them: the one that meant I'd said something interesting, the one that meant I'd said something premature, the one that meant I was asking a question he thought I should answer myself. This was the third kind. It carried a weight that was almost pedagogical, except that Marvenn would never describe himself as teaching me anything. He'd say he was letting me think.

  "You've filled in the justification section," he said, glancing at the form again.

  "Four hundred and fifty words. Three hundred was the target."

  "You wrote more than necessary."

  "The honest answer was longer than the form expected."

  The corner of his mouth moved. He signed. The signature was neat and practiced, the kind of signature that had been written thousands of times across decades of administrative documents, each one carrying exactly the weight it was meant to carry and no more.

  "Let me know how it goes," he said. Which was not the same as let me know if you need help. The distinction sat between us, unexamined.

  The form itself had taken me an hour. The "justification for non-standard classification" section wanted three hundred words explaining why an inscription stylus was necessary for my educational development. I wrote four hundred and fifty, because the honest answer involved formation theory and resonance experiments and mana channel geometry, and trimming it to three hundred would have required omitting things that were true. The "description of item requested" section wanted specifics I could provide (alloy composition, tip diameter, shaft length) and categorizations I couldn't (supply catalog reference number, approved vendor designation). I left those blank and wrote "see maintenance inventory, standard issue" in the margin, which was probably not what the form was designed for.

  Clerk Orveth accepted the form, confirmed Marvenn's signature against a reference list (which he pulled from a different drawer with the same practiced efficiency), and informed me that the request would be routed to Adjunct Marun within two working days.

  "Can I go to Adjunct Marun directly?" I asked.

  "The routing is sequential. Each stage processes in order of receipt."

  "How many requests are ahead of mine?"

  A pause. Not because he didn't know; because the question surprised him slightly. Requests, in his experience, were apparently submitted and then waited for. The idea that a requestor might want to understand the queue was mildly novel. "The current queue for Adjunct Marun contains eleven pending categorization reviews."

  "And the processing rate?"

  "Adjunct Marun reviews categorizations on the second and fourth week of each month."

  Eleven requests. Two review sessions per month. I did the arithmetic and decided not to share the result, which was that my request might not reach the categorization stage for three weeks, leaving only three more weeks for the actual approval process to complete within the six-week standard. The six weeks wasn't a generous estimate. It was a tight one.

  On my way out, I stopped.

  There was a framed document on the wall beside Adjunct Marun's office, posted at a height clearly intended for adult readers, which meant I had to step back to read it from my current stature. The header read: "Procedure for Non-Standard Educational Materials Requests." Below it, a date: thirty-seven years ago. And below that, a notation in smaller text: "Revised from Procedure 14-C, originally established by Deputy Administrator Holbren, Year of the Third Census."

  This story is posted elsewhere by the author. Help them out by reading the authentic version.

  I didn't know the Year of the Third Census offhand. I asked a clerk who was passing through the corridor. She checked a reference chart on the wall (there were many reference charts on the walls of the administrative wing; it was that kind of place) and told me the Third Census had been seventy-two years before the current procedure's revision date.

  So the procedure was thirty-seven years old. Its predecessor, Procedure 14-C, had been in place for seventy-two years before that. The total system for requesting non-standard educational materials had undergone exactly one revision in a hundred and nine years. And the revision, according to the posted document, had changed the routing chain. Not the timeline. The six-week review cycle was original. Deputy Administrator Holbren's original six-week review cycle, running without interruption for a hundred and nine years.

  In my old life, I'd encountered bureaucratic processes that had outlived their original rationale, but usually because nobody had gotten around to reviewing them or because the institutional knowledge of why they existed had been lost. The process had fossilized through neglect. Here, the situation was precisely the opposite. The process had not fossilized; it had been maintained, actively, by people who considered it correct. Deputy Administrator Holbren, if he'd been an Awakening-stage cultivator when he wrote Procedure 14-C, would have had a lifespan of roughly a hundred and twenty years. If he'd been Breathing stage, roughly two hundred. Either way, there was a reasonable chance the man who had originally decided that non-standard educational materials required six weeks of review was still alive. Still walking the halls of this building, or retired from it, or consulting for the office that bore his procedural legacy. And the process he'd created was still running exactly as he'd written it.

  Nobody had changed it because nobody had needed to argue with a dead man's logic. The logic's author was available. If anyone questioned the six-week cycle, Holbren, or someone who'd learned directly from Holbren, could explain. The explanation would be reasonable. The explanation was probably still correct, for the cases the procedure had been designed to handle. And so the procedure endured, not because nobody had noticed it was slow, but because slowness was not a defect in its designers' eyes. Thoroughness was the feature; the six weeks were the cost of thoroughness. If you didn't want to wait six weeks, you could use the standard supply list, which the procedure also governed, and which was updated quarterly by a process that Holbren had also probably written.

  A system whose designers were still alive, and still potentially in the building, and which had not been revised in over a century. Not because it was perfect. Because it was considered finished.

  I was beginning to notice that a lot of things in this world were considered finished, and that the people who considered them finished were often still alive to explain why.

  The six weeks did what six weeks do. They passed.

  I kept working. The worn stylus was still functional enough for the broader channel work; I adjusted my experiments to use geometries that didn't require the precision I'd lost. This was not ideal but it was what I had. Constraints were information, as I'd learned in a different life and a different context: working within a limitation told you things about your process that working without it wouldn't. The broader channels forced me to think about mana flow at a coarser scale, and at coarser scale I noticed structural patterns I'd been too focused on the fine detail to see. A forced step backward that turned out to be useful, the way forced steps backward sometimes are, though I wouldn't have volunteered for it.

  I also learned something about my own impatience, which was sharper than I'd expected. The part of me that had lived a full previous life, that carried decades of experience with institutional delay and red tape, should have been more patient with this. I'd waited years for things in that life. Permits, approvals, grant funding, the particular slow machinery of human bureaucracy in a world where the decision-makers lived seventy or eighty years and still couldn't be bothered to streamline their processes. Here, the decision-makers might live centuries. How was I surprised that the process moved like something designed by people who had all the time in the world?

  But knowing why you're impatient doesn't make you less impatient. It just makes the impatience more articulate.

  Two weeks in, I mentioned the request to Bren during a maintenance round. More observation than complaint. We were in the lower corridor, Bren checking a junction node while I recorded the mana flow readings in the maintenance log, and the topic of tools came up because one of his own styluses had developed a hairline fracture near the tip.

  "I filed a request for a replacement inscription stylus through the educational materials office," I said. "Two weeks ago."

  Bren glanced at me sideways. "Through Orveth."

  "Yes."

  "Six-week cycle."

  "Yes."

  He went back to his junction work. The silence lasted long enough that I'd started writing in the log again before he spoke.

  "What specification?"

  I described the stylus I needed: the alloy, the tip geometry, the gauge. Bren listened without interrupting, which was his standard mode. When I'd finished, he examined the junction sealant he was applying with the focused attention of a man who was thinking about two things at once and choosing to let only one of them show.

  "We have three of those in the maintenance inventory," he said.

  "The maintenance inventory."

  "Maintenance discretionary supplies. Different budget line." He finished with the junction, straightened, wiped his hands on the cloth he kept tucked in his belt. "When I need a tool, I log it under maintenance equipment replacement, pull it from stores, and note the usage in the quarterly maintenance report."

  "How long does that take?"

  "Depends on whether the stores room is staffed. Usually two days. One to log, one to pull."

  Two days. The number sat between us like something obvious that had just become visible. The same item, from the same physical stores (as far as I could tell, the castle's supply stores were a single location served by multiple administrative access points), separated not by distance or availability or scarcity but by which form you filled out and whose desk it crossed. Six weeks through one door, two days through another, and both doors opened onto the same shelf.

  Who else had been waiting six weeks for something sitting on a shelf ten minutes' walk away? How many requests were in Adjunct Marun's queue right now, each one a person who needed a tool or a material, waiting because the form had been routed through the formal channel instead of the operational one?

  "Can you..." I started.

  "I can log it as a maintenance tool replacement," he said. "Because it is. You're assisting on maintenance rounds. The tool is used during maintenance work. The categorization is accurate."

  He said it flatly, the way you state a fact about gravity. There was no guilt in it, no sense of transgression. He was describing a tool he used, in the same tone he'd describe a junction tester or a channel brush. The bypass wasn't a workaround in his mind. It was the appropriate channel for the appropriate need.

  "You've done this before," I said.

  "I do this regularly. The maintenance discretionary budget exists because the formal procurement process is not designed for operational needs. It's designed for planned, predictable, scheduled acquisition. Operational needs are none of those things." He picked up his tool bag. "The maintenance log is the actual resource allocation system for anything that happens on a timeline shorter than six weeks. It has been for as long as I've held this position."

  "How much resource allocation runs through the maintenance log?"

  He considered. "More than it should, by the formal system's standards. The right amount, by any practical measure."

  I thought about this. "Does anyone review the maintenance log?"

  "Quarterly. The household comptroller reviews the total expenditure against the budget line. As long as the total is within tolerance, the individual entries aren't examined."

  "So a single one-line entry in your maintenance log controls more immediate resource allocation than the entire six-week review process."

  "Yes." A beat. "The review process controls allocation for things that can wait six weeks. The maintenance log controls allocation for things that can't. Between the two, most of the castle's actual material needs are covered."

  "Most."

  "There are categories the maintenance log doesn't reach. Large capital purchases. New construction. Anything that requires formal budgetary approval above the discretionary threshold." He started walking toward the next junction point. "Those go through the full process. As they should."

  "But for a stylus."

  "For a stylus, the full process is not designed for a stylus." He didn't look back as he said it. I didn't know yet whether that was frustration or something closer to relief.

  The stylus arrived in two days.

  Bren's maintenance ledger was a thick cloth-bound book he kept in the stores room, its pages dense with thirty years of one-line entries in handwriting that had evolved from carefully legible (the early pages) to efficiently abbreviated (the current ones). The entry for my stylus read: "Replacement inscription tool, fine-gauge, hard alloy (maintenance inventory, standard issue)." Fourteen words. Below it, the stores clerk's initials and a date. Above it, yesterday's entry: "Junction sealant, east wing, node 14-C." Above that: "Illumination crystal, replacement, gallery section 3." Each line a small decision, made and executed, one practitioner's judgment about what the castle needed that day. No committee. No review cycle. No six-week queue. Just Bren, the ledger, and thirty years of knowing what was actually required.

  The stores clerk had pulled the stylus from the same shelf it had always been on, noted the withdrawal against the maintenance budget, and handed it to Bren, who handed it to me during the next morning's round.

  I held it. The alloy was harder than what Garret and I had used, the tip ground to a precision that spoke of professional toolmaking rather than workshop improvisation, the balance in my hand immediately and noticeably superior. A good tool. The kind of tool that should be available to anyone doing inscription work in this castle. And it had been available. For thirty years, at least. To anyone who knew to ask through the right channel.

  I thought about the routing chain. Orveth to Marun to Senrel's division. Six weeks. Every step existed because something had gone wrong, once, and a rule had been added to prevent it from going wrong again. The categorization review existed because materials had been misclassified, at some point in the past, and someone had decided that every non-standard request needed individual categorization. The approval threshold existed because expensive items had been purchased without authorization, at some other point, and someone had set a review floor that, by accident or by design, caught small items along with large ones. The six-week timeline existed because the original review committee had met monthly, and allowing time for the request to arrive, be queued, be reviewed, and be actioned within one committee cycle produced, roughly, six weeks.

  Every step was rational. Every step had been designed by someone solving a specific problem. And the accumulation of rational steps, layered across four hundred years of problem-solving in an institution that never forgot a mistake and rarely forgave one, had produced a system where getting a stylus took six weeks through the formal channel and two days through the maintenance log. Through the patient, relentless accretion of reasonable precautions, each one making the system slightly slower and slightly safer, until the formal path was so encrusted with reasonable precautions that nobody used it for anything urgent.

  The formal path and the actual path had been running in parallel for as long as anyone could remember. The kingdom functioned. Nobody had reconciled them because the kingdom functioned. The bypass was adaptation, the living tissue growing around the calcified bone of formal procedure. And Bren's institutional knowledge, his thirty years of navigating which formal paths matched which operational needs, was the mechanism that kept the gap from becoming a crisis. A practitioner's knowledge, accumulated across decades of daily navigation, filling the space between how the system was designed and how the system was needed.

  My old life had known this pattern. Every institution I'd ever worked in had developed its own version: the formal process and the way things actually got done, running side by side, each sustaining the other in a relationship nobody acknowledged and everybody relied upon. The formal process provided accountability, documentation, the paper trail that proved decisions had been made through proper channels. The informal path provided speed, flexibility, the operational responsiveness that the formal process could not deliver without becoming something it wasn't designed to be. Both were the institution. Neither was the whole story.

  I cancelled the formal request the following week. Went back to Clerk Orveth's office, explained that the item had been obtained through an alternative channel, and asked him to withdraw the form from the queue.

  Orveth looked at me with an expression I couldn't quite read. Not surprise; nothing about a withdrawn request would surprise a man who processed them for a living. Recognition, maybe. As if minor princes withdrawing six-week requests after finding faster paths was a phenomenon he had encountered before, and possibly encountered regularly, and possibly had opinions about that he was professionally obligated not to express.

  "The form has already been forwarded to Adjunct Marun," he said.

  "Can it be withdrawn from Adjunct Marun's queue?"

  "I'll send a withdrawal notice. The original form will be returned to this office for filing. Standard withdrawal processing is five to eight working days."

  "It takes a week to withdraw a request that takes six weeks to approve."

  "The withdrawal process has its own documentation requirements," Orveth said evenly. He had a form for this too, pulled from a different drawer. I watched him fill in the header section: date, requestor name, original request reference number, reason for withdrawal. The reason field had three checkboxes: "Item No Longer Required," "Request Superseded," and "Item Obtained Through Alternative Channel."

  The third checkbox. Someone had anticipated this exact situation frequently enough to give it its own checkbox. Had printed it onto the form, which meant it had been through a design review, which meant at least one other person had approved its inclusion, which meant the phenomenon of people obtaining items through alternative channels while their formal requests were still pending was not merely common but administratively recognized. Documented. Expected. Part of the system.

  The system had a checkbox for its own failure mode. I didn't know whether that was depressing or admirable. Both, probably.

  I thanked Orveth and left. The ledger stayed open behind me, its columns absorbing another entry.

  The corridor of the administrative wing was quiet at that hour, the offices behind their labeled doors processing the castle's operational needs at the pace they had always processed them. The light through the corridor windows was the afternoon light that the castle's original architects had positioned, four hundred years ago, for adequate reading illumination through the working day. It still worked. The procedure on the wall outside Marun's office, a hundred and nine years old, still worked. Orveth's filing system, however many decades it had been running, still worked.

  Everything worked. Everything had been working for a very long time. And because everything worked, nothing changed, and because nothing changed, the gap between the formal system and the actual system grew at the pace of accumulated rules, one well-intentioned restriction at a time, and nobody noticed because Bren, or someone like Bren, had always been there to fill the gap.

  What would happen when Bren wasn't there?

  I walked back to my room with the new stylus in my hand, turning it between my fingers, feeling the balance. A good tool, properly made, obtained through a channel that existed because the formal system couldn't do what it was supposed to do at the speed it was supposed to do it.

  I'd been thinking about workspace. Informally, in the way you think about things you aren't ready to say aloud; I hadn't committed it to paper or mentioned it to Marvenn. But the growing collection of experiments, the inscription pieces accumulating in my room, the documentation projects, the resonance work that needed controlled conditions and space for sustained observation. Sooner or later I was going to need a dedicated space. A workshop of my own, rather than the borrowed corners and undersized tables I'd been making do with.

  A workshop request would not be a non-standard educational materials requisition. It would be a capital allocation, a space designation, a budget line, a staffing consideration. It would cross Adjunct Marun's desk and Senrel's desk and, probably, desks I hadn't even identified yet. The routing chain for a workshop request would make the stylus process look efficient. How many forms? How many signatures? How many weeks, or months, or years of sequential review by offices whose schedules were set by people who would live long enough not to notice the delay?

  And there would be no maintenance log bypass for a room. Bren's discretionary authority extended to consumable tools and materials, the small operational purchases that kept the castle functioning day to day. It did not extend to space allocation. Space was political. Space meant territory, jurisdiction, budget, precedent. A prince requesting dedicated workshop space would be noticed by people whose job was to notice, and the questions they'd ask would not be about inscription tools.

  I set the new stylus on my desk next to the worn one from Garret's forge. Two tools, side by side. The old one, with its broadened tip and workshop-crude balance, made by a twelve-year-old and a smith through trial and error and the willingness to melt down three failed attempts. The new one, with its professional precision and the institutional knowledge required to obtain it. Both functional. Both mine. One acquired through collaboration, one through navigation.

  The afternoon light shifted across the desk. I picked up the new stylus and went back to the inscription piece I'd been working on. The tip held. The channels were clean. The mana flowed smoothly through geometry that matched what I'd intended, line for line, angle for angle.

  But somewhere in the administrative wing, the withdrawal form was beginning its five-to-eight-day journey back through the system. My original request, the one I'd filled out with four hundred and fifty words of honest justification, was working its way backward through a process that didn't have a faster reverse gear. The form would arrive at Orveth's desk, eventually, and he would file it in whatever section handled completed withdrawals, and there would be a record, somewhere in the castle's administrative memory, of a request that had been submitted, processed, bypassed, and cancelled. The full lifecycle of a formal procedure that nobody had used for its intended purpose.

  I wondered if anyone ever read those files. The withdrawal records, the cancelled requests, the forms that had been routed and re-routed and ultimately filed under "Item Obtained Through Alternative Channel." If you read enough of them, a pattern would emerge. A map of the gap between the formal system and the actual one, drawn in the handwriting of every person who'd gone through the front door, found it locked, and walked around to the side.

  The inscription piece in front of me needed my attention. I gave it my attention. The afternoon light moved across the desk and the new stylus held its edge and the work was good, and somewhere in the administrative wing the withdrawal form was making its own slow journey, and the question of who would ever read those files sat in the back of my mind and did not resolve into an answer.