Camie took the news the way she took most things — on the surface, smoothly, with a kind of frictionless acceptance that could mean she genuinely didn't care or that she was filing the information in a compartment you'd never be invited to open.
I'd told her what she needed: delayed personal matter, resolved, operation resumes. I gave her no medical details, no names, no location. What I gave her was a timeline and a task, because Camie operated better when she had something specific to hold.
"The Sekine access point," I said. "You confirmed it's still live."
We were in the back section of a noodle place near Shibuya-ku that had the useful property of being loud enough to make conversation private without trying to be. Camie had a bowl in front of her she'd been working through steadily since I arrived. I had tea I wasn't drinking.
"Still live," she said. "Maintenance rotation changed last week. New window is narrower — down from forty minutes to twenty-two."
"You verified that independently."
She looked at me over her chopsticks. "Yes, Yami. I verified it independently."
The faint edge in her voice was not a problem. Camie was allowed operational pride. What she wasn't allowed was operational pride that bypassed protocol, and so far those had stayed separate categories with her.
"Twenty-two is workable," I said.
"For Toga's part. My part needs the full corridor, which means Toga finishes before minute eight or I'm running a cold approach with no — "
"Toga finishes before minute eight," I said. "That's not a variable."
Camie looked at me for a moment in the way she sometimes did — assessing something she didn't fully say — and then went back to her noodles.
"You look like you got into something," she said.
"I did."
"And we're not discussing it."
"We're not discussing it," I agreed.
She accepted that with the equanimity of someone who'd learned that my information boundaries weren't personal. Or had decided to operate as if they weren't, which amounted to the same thing functionally
Shiro met me at the cache location the following morning.
He was already inside when I arrived, which meant he'd gotten there early or he'd been there for a while, and with Shiro both were equally plausible. He was sitting on an overturned crate with his phone out, doing something I couldn't see the screen of, and he looked up when I came in with the expression he always wore — neutral, waiting, professionally unreadable.
"You're thinner," he said.
"Efficient observation. The Sekine contractor route — current status."
He put the phone away. "Active. My contact confirmed three days ago. The level-two access badge is still associated with the maintenance rotation, and the system hasn't flagged the identifier." He paused. "There's a complication."
I waited.
"The archive sector we're targeting — tier-three — had an internal audit last month. Standard institutional cycle, not a response to anything specific. The audit didn't change the access architecture but it did add a logging layer on documents pulled in the B-wing subsection."
I went through that. "Which subsection specifically."
"B-7 through B-12." He pulled out a folded printout — physical, which I'd trained him toward because digital trails had a specific kind of persistence that paper didn't. "The targets Sera flagged are in B-9 and B-11."
I took the printout and stood with it for a moment.
The logging layer was a problem if we pulled the documents directly. It was not a problem if the documents were accessed, reviewed, and photographed rather than pulled — a distinction that required more time in the sector but left no extraction record.
Twenty-two minute window. Toga on corridor control. Someone on physical access. Someone who could work quickly, under pressure, with fine-motor precision.
I stopped.
The someone couldn't be me. That was the new rule. The rule I'd written in a hospital bed after eleven days of understanding exactly what it cost when the rule didn't exist.
I looked at the printout again. Ran the timing once in my head. Then again, slower.
"The photography work," I said. "It needs to be Camie."
Shiro looked at me. "She hasn't run an archive sector before."
"No. But she's run three office infiltrations and she has better fine-motor control under time pressure than I do at current capacity." I didn't say because my right hand still has a fractured knuckle that hasn't fully resolved. I said at current capacity and let him read what he wanted into it.
He nodded slowly. "She'll need a full brief on document handling protocol."
"I'll run it before the operation." I folded the printout and put it in my jacket. "What else."
"Sera contacted Shiro's relay."
I looked at him.
"Yesterday," he said. "One message. No content. Just a confirmation ping on the channel — she's verifying the relay is still active."
She's testing whether the pause shook you. Puppimil's voice, exact and unhurried, in a building above a tea distributor in Shinjuku.
"Don't respond," I said.
Shiro's expression moved slightly. Not surprise, exactly. More like recalibration. "She'll read silence as — "
"I know what she'll read it as. Don't respond."
He filed that without further discussion. Which was one of Shiro's better qualities: he could disagree with a decision and implement it cleanly anyway, in the understanding that disagreement got voiced once and then either changed the call or didn't.
---
I spent the afternoon on Puppeteer.
Not operational prep. Not target research. Just the quirk itself, in the empty back room of the cache location with the lights low and nothing to practice on except the quality of my own attention.
This was something Puppimil had said almost in passing, near the end of the meeting — you understand the tool architecturally but you haven't mapped its internal range. She'd said it the way she said things that she wasn't going to elaborate on, which meant she expected me to elaborate on it myself.
Puppeteer worked by redirecting attention. That was the surface description and it was accurate but it was like describing a surgical instrument as a thing that cuts — technically correct and operationally insufficient.
What I knew from practice: attention had texture. The way a person was paying attention to something was different from the fact that they were paying attention to it, and Puppeteer could work on both the direction and the texture simultaneously, which was what made it useful at distances where the target wasn't consciously aware of the manipulation.
What I knew from the Hando incident: close range introduced a feedback element I hadn't accounted for. A trained subject at high alertness generated a kind of attentional density that had weight — it pushed back, not through a countermeasure but through sheer force of presence. Like trying to redirect a current that was moving faster than I'd modeled.
What I didn't know: where the internal ceiling was. Whether the feedback was a ceiling or a threshold. Whether the cost curve was linear or whether there were drop-off points I hadn't hit yet because I'd stayed in the safe operating range for so long I'd mistaken the range for the full map.
I sat with the quirk in the quiet room and ran it at nothing — just the activation, the attention-sense that came with it, the particular quality of perception that made the world feel like a map of where everyone's focus was resting — and I tried to feel the edges of it without pushing past them.
It was, I thought, a little like pressing against the inside of a container you were trying to understand the shape of. You found the resistance. You noted where it started. You didn't push through it — not yet — but you mapped its surface.
Slow work. Not satisfying in the way operational work was satisfying.
But the ambition problem was separate from the environment problem. Puppimil had said so. And the ambition problem required knowing the actual edges of what I was working with, not the edges I'd assumed from a sample size that was, I now understood, embarrassingly small.
I sat with that for another long moment before I pulled out my phone and filed what I'd observed.
Feedack at close range is density-based, not skill-based. May be consistent across hero-tier subjects or may be individual. Sample size: one. Insufficient. Texture manipulation distinct from direction manipulation — separate cost curve? Unclear. Range assumption: needs real testing. Controlled environment.
I looked at the last line.
Controlled environment — Puppimil's words, her offer, the condition attached to access to the upper floor.
I hadn't agreed to the arrangement explicitly. I'd left without agreeing to anything except that I was watching and she was watching and the test was still running.
But I was already doing the homework.
I closed the note and went back to operational prep.
---
Three days before the archive job, Sera moved.
Not a ping this time. A message on the relay, two lines:
Timeline status. Meridian's position adjusts end of month.
End of month was nine days. Which meant the window for the tier-three delivery was not open-ended, which I'd suspected but hadn't confirmed.
I read the message twice. Let it sit.
Meridian's position adjusts. That was new language. Sera didn't use operational terminology casually — she used it precisely. Which meant Meridian was not a person but a position, and the position was adjusting, and the adjustment had a deadline attached.
I ran through what that implied. A principal moving. A window closing. A delivery that had to happen before the relocation, or the chain broke.
I'd known there was pressure. Now I knew the shape of it.
I sat with that for a while before I wrote back. One line:
Delivery confirmed before end of month.
Then I called a final pre-operation brief with Toga and Camie for the following morning.
---
The brief ran ninety minutes.
I'd rebuilt the briefing structure during recovery — cleaner sequencing, explicit buffer protocols, contingency trees that went three branches deep instead of one. Not because the old structure had failed but because I'd had eleven days and a fractured hand and I was going to use them for something.
Camie ran the document photography protocol without complaint, asked two precise questions about the logging layer and accepted my answers, and spent the last twenty minutes of the brief going over the sector map in the focused way she had that made her look like she was reading something visible to only her.
Toga didn't ask questions. She absorbed the brief in her particular way — present, attentive, processing through whatever internal system she used — and at the end she said, "Minute eight. Confirmed," and that was the totality of her response.
After they left I sat alone in the cache room with the operation laid out in my head and ran it forward in the honest register.
The risk points: the logging layer, the narrowed window, Camie's first archive sector, Shiro's access badge holding through the operation window, Sera's end-of-month pressure creating timeline stress that could push decisions in ways I hadn't modeled.
The controls: buffer assets, distance protocol, layered exits, the new three-branch contingency structure.
The honest assessment: there were six things that could go wrong and I had solid controls for four of them, partial controls for one, and the logging layer was the one I was holding with the least certainty.
If the logging layer flagged an access pattern even without a document pull — if the audit had added presence detection as well as extraction detection — then the sector visit itself was a record.
I sat with that for a while.
The problem wasn't the logging layer itself. The problem was that I didn't know what kind of logging layer it was. There was a difference between documenting what was removed and documenting who was present, and I didn't know which one the audit had installed.
I could ask Puppimil. She'd have access to the specification or she'd know how to get it. But that meant walking into the Shinjuku building with an open question instead of a finished operation, and I wasn't sure yet what that signaled.
I sat with that too.
Then I made a different call.
I added a seventh step to the brief: Camie would spend no more than eleven minutes in the sector regardless of completion status. Partial documentation on the first pass was not a failed operation. It was a partial operation with a clean exit. I could build around partial.
What I couldn't build around was a logged sector access that led back through the Sekine contractor chain to Shiro, to the relay, and outward.
Eleven minutes. Hard limit. New rule.
I added it to the operational document and sent the update to Camie's relay.
Her response came back in four minutes:
Understood. Eleven.
---
Two days before the operation, I went back to the Shinjuku building.
Not because Puppimil had asked me to. Because the question I'd been sitting on had matured from something I don't know to something I need to know before I run this operation, and I'd learned enough about how she worked to understand that the distinction mattered.
She opened the door before I knocked this time.
"You're early," she said.
"The operation is in two days. I have a specific question."
She stepped aside. I went in. The room looked the same — documents, maps, maintained surfaces. She went back to the center table and I remained standing.
"The logging layer on the archive sector," I said. "Presence detection versus extraction detection. Is there a way to determine which type was implemented without running the operation and finding out the hard way."
She looked at me for a moment. Not evaluating, I thought. Letting the question settle.
"Sector B-7 through B-12," I said. "Tier-three municipal archive. Internal audit last month."
She turned to one of the bookshelves, ran her finger along a row of physical files, pulled one, opened it. Read for thirty seconds.
"Extraction detection," she said. "The audit budget didn't cover presence upgrades. Presence architecture costs roughly four times extraction architecture and requires infrastructure the municipal system doesn't currently support."
She closed the file. Put it back.
I stood there with the relief I wasn't going to perform and the irritation that I could have walked in here two days ago and asked this directly. But that, I understood, was the wrong read. I hadn't known it was the right question two days ago. I'd needed to build the operation to the point where the uncertainty became specific.
"You could have told me to ask," I said.
"Yes," she said. "But you didn't know you needed to ask until you'd built the operation carefully enough to identify the precise risk point." She picked up her pen. "That's the difference between bringing me a question and bringing me a good question."
I looked at her.
"The operation's ready," she said. Not a question.
"Yes."
"Run it clean. Come back with the result." She made a notation in a margin. "Then we'll talk about first real money. What you've been earning — that's apprentice rate. Survival income. What the upper floor pays is different in kind, not just in quantity."
I moved toward the door.
"Yami."
I stopped.
"Eleven-minute limit on the sector," she said. Without looking up from the document. "That's the right call."
I left without asking how she knew.
I was beginning to understand that with Puppimil, the answer to how does she know was almost never the useful question
The useful question was always what she was doing with it.
And the answer to that one, I was slowly concluding, was: building something.
Whether I was the architect of that something or one of its components was still, technically, an open variable.
I walked down the herbal-scented stairwell and back into the city and decided that the way to resolve that particular variable was to keep being useful enough that the distinction stopped mattering.
The operation was in two days.
I had work to do.
Visit My Patreon and read upto 30+ Advanced chapters in advance to support the story
My Pat reon:ht tps://patreon.com/eyestapler
Thanks for reading