The fourth bank job went wrong in a way that didn't look wrong until afterward.
Not during. During, it was clean. Nori's interference held. Riku handled the physical presence without unnecessary movement. Daigo managed the lobby — two staff members and one early customer, all redirected through conversation and mild misdirection that didn't require anything as expensive as quirk use. Extraction was fourteen minutes. Routing was clean.
We were out. The money cleared.
Three days later Hatsue came to me with a printout.
A reconstruction security circular — wider distribution than the newsletter, maybe two thousand readers, mostly insurance adjusters and private risk consultants and the kind of ex-hero support staff who had moved into corporate security roles after the licensing restructure. It noted a pattern across six financial institution incidents in the greater Musutafu area over a two-month window.
Six.
Not three. Six.
I read the circular twice.
The three jobs I hadn't run were not us. But they fit our structural signature closely enough that someone doing pattern analysis had grouped them together. We were being credited with work that wasn't ours, which meant the profile attached to our actual work was getting denser and more visible through no action of our own.
I set the printout down.
"Where did the other three come from?" I asked.
Hatsue said she didn't know yet. She was working on it.
I told her to make it the priority.
---
The answer came back in two days.
The other three incidents were the work of a two-person operation — a man and a woman, neither with confirmed quirk profiles in the available records, running legacy-infrastructure financial targets in the same corridor we had been using before I pulled us out of it.
They weren't affiliated with us. They had no connection to our network. They had simply identified the same target class independently, probably through the same kind of pre-war infrastructure mapping that Nori had done.
Which meant the corridor had been independently identified as viable by at least two separate operators.
Which meant the corridor was now a liability regardless of whether we used it again.
I told Nori to decommission all the infrastructure maps for that target class and rebuild the assessment criteria from different entry points.
He took it without complaint, which was not how he would have taken it three weeks ago. He was learning the difference between losing work and avoiding a mistake.
That was worth noting.
---
The friction problem was subtler than the circular.
The circular was external — a risk to be managed, a corridor to be abandoned, a process to be adjusted. I knew how to handle external friction.
What I was dealing with inside the team was harder to name.
It wasn't conflict. The team was functionally aligned. Everyone was performing at or above the standard I had set. The jobs ran. The consultancy was growing. The money was moving cleanly.
But there was a texture to the interactions that had changed in the past two weeks, and I had been tracking it without yet knowing what to call it.
Hatsue had started occasionally answering questions I hadn't asked. Not overstepping — the answers were always relevant. But anticipating where I was going. Which either meant she was developing a strong model of how I thought, or she was testing whether I would accept initiative.
Nori had started suggesting variations during job planning rather than waiting to be asked. Again — useful. Again — something to watch.
Daigo said nothing differently. But he had been reporting observations about the other team members more frequently, with more unsolicited detail.
And Riku had been quiet in a way that was not his normal quietness. His normal quietness was the quietness of a person who didn't have much to say and was comfortable with that. This was the quietness of a person who had something to say and was deciding whether to say it.
I let it sit for a week.
Then I asked him directly, alone, before a Thursday briefing.
"You've been carrying something. What is it."
Not a question. A statement that gave him permission.
He looked at his hands for a moment.
"The woman I talked to," he said. "She asked me again."
"About what."
"About what I do."
I waited.
"She's not connected to anyone. She's just — she's curious about me. As a person." He said the last part with the specific discomfort of a man who was not used to being curious about himself. "I haven't told her anything. But she's going to ask again."
I thought about it.
"What do you want to do?" I said.
He looked up, mildly surprised. I had not asked him that before.
"I don't know," he said. "I thought you should know first."
I told him I appreciated that.
Then I told him to think about what he actually wanted and come back to me before the next job.
He nodded and went into the briefing room.
I stood in the hallway for thirty seconds and thought about what I had just done.
I had given him permission to want something. That was not the same as giving him permission to act on it. But it was closer to that than my previous management approach, which had been to treat the wants of team members as variables to be accounted for rather than inputs to be engaged.
The question was whether that was the right adjustment or a drift toward something less controlled.
I didn't have the answer yet.
I went into the briefing.
---
Shinkō Data had its first difficult client.
A logistics coordinator — not the same one from the medical supply job, a new referral — came in with a request for an assessment of ward-level contract decision patterns in the Koto district. Standard framing. The kind of question we had been answering for two months.
But the specific documents he wanted cross-referenced were not public reconstruction records.
They were internal procurement variance reports — the kind that existed only inside ward office systems and were not supposed to circulate.
I looked at him across the table in the clean office Hatsue had set up for client consultations — separate from the operations space, legitimately furnished, looked exactly like what it was supposed to look like.
"Those documents aren't in our data set," I said.
"I was told you could find things that weren't in normal data sets," he said.
The referral chain had leaked something about our methodology.
I noted that while keeping my expression neutral.
"Whoever told you that oversimplified," I said. "We work with publicly accessible reconstruction data and proprietary analytical methods. We don't acquire non-public documents."
He looked at me for a moment.
"What if the procurement variance reports were provided to you. By a source. You just do the analysis."
I thought about what kind of person sent a source with non-public procurement documents to a consultancy that was three months old and had no formal institutional standing.
Two possibilities.
One: he was a legitimate corporate client with a grey-area information acquisition he wanted laundered through a respectable-seeming analytical layer. Moderately risky for us. The exposure would be his if the documents were traced.
Two: he was a probe — someone checking whether Shinkō Data was willing to handle irregular material, either to exploit that willingness later or to establish a compromising relationship.
I looked at his posture. The way he was sitting. The specific cadence of his follow-up question.
He was not comfortable.
Someone comfortable with irregular requests didn't show it this way — he was performing a kind of casual certainty that the underlying discomfort was leaking through. That was either inexperience or constructed behavior designed to look like inexperience.
I decided the calculation was not favorable regardless of which it was.
"We're not the right firm for that," I said pleasantly. "But I can give you a referral for a firm that handles forensic document analysis for authorized parties, if the documents in question have a legitimate provenance."
He took the referral.
He didn't call the number I gave him, which was a real firm that would ask for documentation we both knew he didn't have.
I told Hatsue to flag the referral chain he had come through and run a light assessment on whether the source had been talking loosely or had been deliberately positioned.
She came back the next day.
The source had been talking loosely. Probably. The referral had passed through three conversations before reaching this client, and somewhere in that chain someone had implied capabilities we hadn't claimed.
I tightened the referral language. Added a specific framing that clients were supposed to use when recommending us. Made it clear that imprecise description of our services created expectations we would decline to meet.
The problem was managed.
But it reminded me that legitimate cover had its own friction — the friction of a public-facing operation that people would talk about in ways I couldn't fully control.
---
The Yaoyorozu thread moved.
Not because I had done anything. Because the board minutes from the most recent materials allocation meeting had been published in a reconstruction transparency report, and Ikeda's formal recommendation in that meeting had contained a citation error.
Small. Easily missed.
He had cited a materials assessment report dated three weeks earlier — but the assessment methodology it described had not existed three weeks earlier. It had been developed and distributed internally only ten days before the meeting.
Which meant the report had been backdated.
Which meant Ikeda had made a mistake — had cited a document that hadn't existed yet as if it had informed a previous decision — in a transparency report that was technically public.
I read it four times.
Then I archived the original transparency report in three separate locations.
Backdated documentation in a procurement context was not just administrative sloppiness. It was evidence of retroactive justification — someone had made a decision, then constructed the paper trail to make it look like the decision had been properly informed.
Ikeda had left a seam.
Not a large seam. Not one that most people would find or know what to do with.
But I was not most people and I knew exactly what to do with seams.
I didn't move on it immediately. The seam was more valuable as a held piece of information than as an immediate action. Ikeda didn't know I had it. As long as he didn't know, it was leverage. The moment I used it, it became something he would close around.
I noted the date, the document reference, the citation error, and the ten-day impossibility window.
Then I labeled the note *Hold. Do not activate.*
---
That evening I thought about the operation as a whole.
Three months since the Kanda office. Two since Suidobashi. The team was stable. The legitimate income was growing. The criminal income was varied, low-profile, and building toward something I hadn't fully defined yet.
I thought about where the ceiling was.
Not the ceiling Kuroda had described — the contractor ceiling, the one I had already passed. A different ceiling. The one above the second economy, above the Lattice, above the level where people like Puppimil operated.
I didn't know what was there.
I knew it existed because everything I had mapped had space above it that I couldn't see into.
And I knew that what was happening around Yaoyorozu — the scale of it, the care with which it had been constructed, the principal I still couldn't name — was adjacent to that ceiling.
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