The legitimate income came from Hatsue's idea, which I approved because it was good, not because it was hers.
She had noticed something I had catalogued but not yet acted on: the reconstruction economy had created an enormous demand for information brokerage. Not illegal information — just scarce information. Which contractors were solvent. Which supply chains were functional. Which ward offices were processing permits on normal timelines versus the ones that had three-month backlogs because two of their senior staff had died in the war and nobody had filled the positions yet.
People needed that information and had no clean way to get it.
"We already have most of it," she said. "From the contractor credential work. We just never monetized the data itself."
I looked at her across the table in the Suidobashi office.
"Walk me through the liability surface," I said.
She did. Fifteen minutes, no notes, clean logical structure. The information we had was acquired through methods that couldn't be disclosed, but the information itself was not classified. Selling it through a clean front — a reconstruction consultancy, registered properly, with a legitimate business address — would generate taxable income, create a paper trail that supported a normal professional identity, and provide cover explanation for money that currently had no source.
I told her it was good work.
She didn't react visibly. That was also good work.
---
The consultancy was registered as **Shinkō Data Services** within the week.
The registration process was tedious in the way that legitimate things are tedious — forms, a registered address, a bank account, a listed director who was not me but was a name connected to a document chain that Daigo had built over the previous month for exactly this kind of use. The listed director had a history, a tax record, and a residential address in Nerima ward.
He did not exist.
But his paperwork was better than most people's real paperwork, and in the reconstruction era, paperwork was what institutions ran on because the physical infrastructure for anything more sophisticated was still being rebuilt.
First client came in eleven days after registration.
A mid-size concrete supplier trying to assess which ward procurement offices were processing emergency materials contracts on accelerated timelines. Standard reconstruction problem — the supplier had capacity but no visibility into where demand was actually moving.
Hatsue pulled the assessment in four hours. I reviewed it, removed three data points that would have been obviously irregular to anyone who thought hard about how we obtained them, and replaced them with inference language that arrived at the same conclusions through stated reasoning.
We charged forty thousand yen.
The supplier paid without negotiation.
I sat with that number for a moment.
Forty thousand yen. Clean. Taxable. Explainable.
The amount was not impressive. The cleanliness was.
---
The third bank job was where I made my first real mistake in six weeks.
Not an operational mistake. A judgment mistake, which is harder to catch because it doesn't announce itself until after.
The target was a branch in Itabashi — pre-war infrastructure, same vulnerability class as the Adachi jobs, Nori had already mapped it. The job was structured identically to the second Adachi job, which had been the cleanest run we'd done.
The mistake was that I ran it identically.
Same timing window. Same role distribution. Same extraction routing.
It worked. The job succeeded. We cleared the branch in twelve minutes and the extraction was clean.
But three days later Hatsue brought me a note: a low-circulation reconstruction security newsletter — the kind written by ex-hero support staff who had moved into private consulting — had published a brief observation. Three financial institution incidents in the Adachi-Itabashi corridor over forty days. Possibly coincidental. Possibly indicating a localized opportunistic actor with knowledge of legacy infrastructure.
The newsletter had a readership of maybe four hundred people.
Four hundred was enough.
I read the note twice. Then I told Hatsue to archive it under active monitoring and called the team in.
"We are not running another job in that corridor," I said. "Not for ninety days. We are also varying the structural template. Same job run the same way three times is a pattern, and patterns are the only thing investigators have to work with when they have no leads."
Nori said, "We weren't identified."
"We weren't identified *yet*," I said. "Those are different sentences."
He wrote it down. That was new. He had started taking notes two weeks ago without being told to. I had noticed and said nothing, because acknowledging it would have made it a performance instead of a habit.
---
I spent the following week varying the templates.
Not randomly — random variation is as identifiable as consistency, because genuine random behavior has a texture that humans can detect even when they can't articulate it. I varied the templates according to a rotation logic that was internally consistent but externally discontinuous. Different entry timing, different quirk load distribution, different extraction architecture.
The pattern would still exist.
But it would require a much more sophisticated analyst to find it, and those analysts were not working reconstruction security newsletters.
Not yet.
---
The love scam infrastructure had its first complication.
One of Daigo's personas — the one he called Takahashi, a grief-widowed structural engineer from Chiba — had developed a target who was responding not with financial openness but with reciprocal emotional investment. The target, a fifty-three-year-old woman who had lost her husband and two colleagues in a warehouse collapse during the war, had begun talking about meeting in person.
Daigo brought it to me.
"She's asking for a video call first," he said. "Real-time. I can't run a static persona through real-time."
I considered the options.
Escalate to a deepfake video infrastructure, which we didn't have. Stall and risk losing the target's investment. Or gracefully resolve the persona — have Takahashi go quiet, then surface briefly with an excuse that cooled the connection without complete severance, preserving the option of a later financial ask.
I chose the third.
Daigo executed it cleanly. Takahashi became temporarily unreachable due to a work deployment complication. The target was disappointed but not alarmed. The connection dropped to low-maintenance contact.
Afterward I noted: *we have a ceiling on relationship depth. Do not exceed it until we have real-time persona capability or a better resolution method.*
Then I noted something else, in a separate column I had started keeping for things I didn't want to lose:
*She talked about her husband like he was still in the room.*
I didn't do anything with that observation.
I just kept it.
---
The Yaoyorozu file was getting denser.
I had started pulling thread-level detail rather than pattern-level. Specific meeting attendance, specific document timestamps, the subtle timing differences between when her assessment input appeared in board materials versus when the formal recommendations were filed.
There was a delay. Consistent. Four to seven days between her input and the formal recommendation.
That delay was being used to launder her analysis through an intermediary voice — someone taking her conclusions and re-presenting them as their own institutional assessment, with her work visible enough to be utilized but not visible enough to be attributed.
She was being cited in the body of documents but not in the authorship.
I pulled the name that appeared most frequently as the formal recommending officer.
**Ikeda Sho.** Forty-four. Former hero agency logistics director, reconstruction advisory role, strong political alignment with the current ward governance structure. He had moved from hero support bureaucracy to reconstruction bureaucracy without a visible gap, which meant he had known the transition was coming or had been positioned for it.
I opened a new file. Labeled it *Ikeda — Structural Relationship, YMM.*
Then I sat back and looked at the two files side by side.
Yaoyorozu was the capability. Ikeda was the extraction mechanism. Somewhere above Ikeda was a principal who had identified the value of a Creation quirk user and built a quiet structure to utilize it.
The principal I couldn't name yet.
But I could map the structure.
And structures had entry points.
---
The thing I kept coming back to was the question I had written at the top of the Phase One document.
*What does she think she's building?*
I had three weeks of observation now and no clean answer.
She appeared, from the outside, to be exactly what she presented: a deactivated hero applying her capability toward civilian reconstruction, working within institutional frameworks because that was the only available framework. Dutiful. Organized. Precise.
But twice in the board minutes I had flagged a specific pattern: she had asked questions about contract recipient accountability. Not technical questions. Structural questions. *Who verifies that the materials are being deployed where they're intended?* and *What is the audit mechanism for emergency procurement exceptions?*
Both questions had been addressed briefly and moved past.
She had not pushed.
But she had asked.
That told me something.
She was not entirely unaware. She was uncomfortable with something she couldn't yet name, or couldn't yet prove, or didn't yet feel she had the standing to formally challenge.
That was not the profile of someone fully captured by the system using her.
That was the profile of someone waiting to understand what was wrong.
---
I wrote a single line under the question.
*She thinks she's building something honest. She is starting to suspect she isn't.*
Then below that, in the shorthand I used when I was working through a long-range problem:
*Timing: she needs to reach the suspicion on her own. Contact before that point closes the window. Contact too long after — she builds her own response and doesn't need an outside frame.*
*Window: not open yet. Monitor for the acceleration event.*
---
Daigo told me, offhandedly, at the end of a Friday debrief, that Riku had started talking about the operation to someone outside the team.
Not specifics. Not actionable intelligence. Just the fact of being part of something. The vague pride of a man who had previously had no structure to belong to.
I thanked Daigo and asked who Riku had been talking to.
A woman he was seeing. No connections to anyone in our operational adjacency. Low risk on the surface.
I called Riku in separately the next morning.
I didn't express anger. Anger would have made him defensive and defensive people lie to manage the emotion in the room rather than to give you accurate information.
I said, "I need to ask you something directly and I need a direct answer."
He gave me one.
It was low-level. A sentence, maybe two, in a conversation that had been about something else. The woman in question had shown no particular interest.
I told him: "I understand why it happened. I'm not angry. But it doesn't happen again. Not a sentence. Not an implication. Nothing."
He said yes.
I believed him.
But I also began thinking about what Daigo's role in the team actually was, under the surface of the job functions I had assigned him.
He noticed things about other people. He reported them upward. He was precise about it.
That was useful.
It was also something I needed to hold carefully, because a person who watches the team for the leader will eventually start watching the leader for someone else, if the price is right or the reason is compelling enough.
I made a note.
*Daigo: asset and variable. Do not forget the second part.*
---
By the end of the week the operation had three income streams, a functional cover identity, a secondary location, and a problem I hadn't fully named yet walking around outside the Kanda Street office in an old jacket.
The unnamed man still hadn't reappeared.
I had stopped expecting him to reappear and started expecting him to have already done what he came to do.
What that was, I didn't know.
What it would cost me, I couldn't calculate yet
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