Three weeks.

Soma turned over the floorboard before the sky had fully brightened, looked at the list of notations in the weak light from the dying kitchen coals, and began his assessment.

Not by feeling. By data.

Bridge footwork technique — stable. He knew this because three days ago he had crossed it while thinking about something entirely different and still made no sound. When a skill no longer required conscious attention to function, it had moved into muscle memory. That was the threshold he had been aiming for.

Low center of gravity — progressing. Not yet natural — he still had to remind himself at the start of each movement. But the duration he could maintain the correct posture had increased from a few minutes to nearly half an hour before his quadriceps began objecting.

Peripheral vision — difficult to assess. He was paying more attention to movement at the edges of his field of vision than before. Whether that was real improvement or the illusion of someone focused on a thing, he could not yet determine.

He set the charcoal down. Replaced the floorboard. Stood up.

Three weeks. Not bad. Not enough.

Hiroshi was still asleep. Soma had fifteen minutes before he needed to be at the kitchen fire. He stepped outside into the last of the dark and let his mind run in whatever direction it wanted.

And something was wrong.

He stood at the edge of the irrigation ditch and felt the wrongness without being able to name it immediately. A restlessness in the way he had been approaching his practice. Like an itch in a location he could not reach.

It took him until the first birds began to call to understand what it was.

He had been learning what without understanding why.

---

He knew what: bridge footwork, low center of gravity, peripheral vision. He had copied observable technique from the genin he had watched and worked backward to the physics.

But he had never asked the deeper question.

Why did these specific techniques exist? What problem had the people who developed them been trying to solve? What was the solution space available to a shinobi — and which parts of that space had been explored, which left empty?

He stood at the ditch bank in the early morning and let the question open up.

Shinobi solve what problem?

Not combat — not primarily. Samurai solved combat. Shinobi were hired for things samurai could not or would not do: infiltration, information, removal of obstacles without public acknowledgment. The fundamental constraint was not strength but invisibility. The absence of evidence that anything had happened at all.

So the solution space was: how does a person move through an environment occupied by other people without those people being certain of that person's presence?

Every technique he had observed was a response to a specific sub-problem within that space.

He spent the entire morning testing the idea.

He walked through every type of terrain around the village — the dirt road, the paddy banks, the sparse woods to the north, the low rooftops of the abandoned storage sheds to the east — and asked himself the same question about each: if I needed to move through here without being detected, what problem would I face?

Dirt road: hard surface, visible footprints, dust rises with heavy impact. Problem: minimize pressure at contact points, avoid disturbing surface material. Solution space: light step, choose grassed edges over bare earth, walk in existing footprints when available.

Paddy bank: soft soil, unstable, narrow raised edges. Problem: balance on unstable narrow surface without creating sound of water or settling earth. Solution space: low center of gravity, test each step before committing weight — identical to rock climbing technique he had read about in a magazine.

Sparse woods: uneven light, multiple obstacles at eye level, dry leaves and branches on ground. Problem: move through three-dimensional irregular space without making contact with sound-generating objects. Solution space: look far ahead to plan path, probe foot before placing weight, keep hands close to body to avoid sweeping branches.

He was no longer copying. He was constructing.

---

By mid-morning, his body was tired from the terrain walk.

He had discovered the first genuine limit of the afternoon: this eight-year-old body had stamina built from farmwork — good horizontal endurance — but the concentrated fine motor control he was demanding exhausted it faster than he expected.

Not a problem of technique. A problem of foundation physiology.

Progressive overload, Kenji noted. Start at the lowest sustainable threshold and increase incrementally according to what the body can absorb. It will take years.

He accepted that without hesitation.

He stopped at the abandoned storage shed on the eastern edge of the village — low roof, still mostly intact on the right side, earthen walls cracked and patched with moss. No one had used it for three years according to Soma's memories. The previous owner had died without an heir.

Soma stood outside it and looked up at the roof.

Height approximately two and a half meters from ground level. Roof pitch roughly thirty degrees — not too steep. Straw decomposed unevenly in patches, some sections still firm. The earthen wall could serve as a climbing surface if the right grip points were found.

Kenji had calculated this the previous evening.

Frictional force between palm and damp earthen wall: sufficient for short-duration grip if contact angle was correct. Force required to lift body weight: feasible with current arm and leg strength given proper fulcrum points and timing. Three points of contact minimum at all times — basic rock climbing principle.

He placed his hands on the wall and began.

First grip point — solid. Pulled himself up, feet searching for purchase on the wall surface. Stable. Right hand reached for the next point —

Slip.

Not a gradual slip — sudden, without warning. The wall at the point he had gripped had a thin dry layer over a damp interior — his hand broke through the dry layer and lost all friction in a fraction of a second.

He fell.

Not dangerously — he caught himself instinctively, right elbow hitting the ground first. Not serious. But painful and surprising in the way laboratory prediction never fully prepared for field reality.

Soma sat on the ground, looked at his scraped elbow, and said nothing for a moment.

Then Kenji started analyzing.

Where did the model fail?

He had calculated static friction — the force holding a stationary object in place. But when moving, the force on the grip point was not static but dynamic — larger and directionally variable as the body shifted. He had also failed to account for surface heterogeneity. The earthen wall was not a uniform surface. His model had assumed consistency that reality never provided.

In other words: he had calculated for a textbook object, not for an actual child climbing an actual wall.

The framework was not wrong in principle. The model needed to be corrected.

He stood up, brushed dirt from his knees, and looked up at the wall again.

Adjustment: add a safety factor for surface non-uniformity. Test each grip point with partial weight before trusting it fully. Do not transfer weight sharply — always have the third contact point stable before moving the first.

He tried again.

This time slower — each grip point tested with a fraction of his weight before committing. Speed was half the first attempt. Not efficient. But he reached a third of the wall height without slipping.

Stop. Descend. Note. Try again.

Three more attempts — two minor slips, both less severe than the first because he had adjusted his catch. By the fifth attempt he had reached the roof edge and sat on it, looking down at the village.

No feeling of victory. Just a note: adjusted model performs better. Needs more data from different surfaces.

He sat on the roof a few minutes longer — not for technical reasons but because from here he could see the whole village in a way he never had from below.

Small. This village was smaller than he had thought when looking up at it from the ground.

---

That afternoon he returned to the eastern edge of the village.

Not by accident this time. He went deliberately, looking for traces of the person he had seen the week before, intending to read more from what remained.

He started from the spot where he had been sitting when he observed: the low shrub about twenty steps from the reed bed. Looked toward the reeds and worked backward: if I were the person standing there, which direction would I have arrived from and which direction would I have departed?

He circled the reed bed in a half-arc, eyes on the ground.

First trace found after about five minutes — a slight compression in the grass, not clear enough for an ordinary person to notice but distinct enough to distinguish from grass displaced by wind or small animals. The shape of the compression suggested the ball of a foot placed forward, weight distributed heel to toe — the same stepping pattern he had directly observed.

He followed the direction the ball of the foot indicated.

Next: a low branch, chest height, bent at an angle that was not natural — not wind, not animals, but a hand that had gripped it to maintain balance while stepping over waterlogged ground. The branch was bent at the optimal point to absorb load without breaking — the person who had done this had known exactly how much force to apply.

Further on: a faint handprint on a low earthen wall, four fingers visible, no thumb — the person had gripped from above rather than from below. Soma looked at the wall and understood immediately: this wall was low, requiring no upward pull — only a balance point while stepping over. Gripping from above was faster and left less forceful evidence than gripping from below.

Small detail. But someone had thought about it.

He followed the traces for another two hundred steps eastward — out of the village edge, through a stretch of overgrown wasteland, and into a narrow stand of thin-trunked trees he had never had reason to approach.

He stopped at the edge of the trees.

Looked in.

No unusual sounds — birds still calling, wind moving normally through leaves, no sign of active presence. He stood for several more minutes and allowed his peripheral vision to work before stepping forward.

Nothing moved.

He stepped into the trees.

---

The first thing he found was the fire pit.

Not immediately — he had walked past it before Kenji's attention pulled him back: the earth at that spot was a shade darker than the surroundings, grass growth thinner, and there was a small circle of stones that looked natural at a glance but had been arranged with too much regularity.

He crouched beside the stone circle.

Old ash — at least a week, possibly more. Covered with earth and leaf matter carefully, the way someone concerned about concealment would cover it rather than the way someone who merely wanted to tidy up would. This was a person who did not want anyone to know they had cooked here.

Around the pit, the ground had been cleared of dry leaves within a radius of about two meters — reducing the risk of fire spreading and reducing the sound of movement in the dark. Defensive thinking even in daily living.

Then he saw the cord.

Stretched between two tree trunks, roughly chest height on an adult — not chest height for a child, he had to look for it specifically. Dark brown cord, blending with the bark color, nearly invisible unless you were searching. Tied to both trunks with simple but secure knots.

A warning snare.

Not designed to cause injury — cord at that height would only catch against a passing person's body, creating noise or movement that would disturb someone sleeping nearby. Sufficient to wake anyone sensitive enough to have thought to set it.

Soma recognized it and did not touch it.

He looked at the entire area from where he stood — the fire pit, the cleared ground, the warning cord — and read what he saw the way Kenji read a technical drawing.

Someone lives here. That person is careful. That person does not want to be found.

He turned around and left by exactly the path he had entered, careful not to leave new traces on top of what was already there.

When he reached the edge of the trees, he stopped once and looked back.

The shadows between the trunks showed nothing moving.

There is someone here, he thought. That person is careful. That person knows things he needed to know.

He walked back toward the village in the late afternoon light and did not look back again.

---

That evening Soma lifted the floorboard for the second time that day.

Looked at the list. Crossed out some old questions, added new ones. Adjusted the wall-climbing framework for what he had learned that morning. Recorded the location of the forest and what he had observed.

Then he stopped at the third column — unanswered questions.

And faced a truth he had been avoiding.

He could teach himself to the limit of solo practice. But there were things that could not be learned alone. Real combat. Reflexes under pressure. How the body responded when someone real was genuinely trying to cause real harm. Nothing he practiced alone could prepare him for that. Those things only came from contact with another person — and he had no way of creating that contact safely.

This variable had no solution yet.

He did not try to force one. He left the question on the list — an open question was information, not failure. It reminded him there was a genuine limit to how he was currently progressing, and that limit would need to be addressed in a way he had not yet found.

Perhaps the person in the forest was part of the answer.

Perhaps not.

He replaced the floorboard, set down the charcoal, and lay back.

More observation tomorrow. More practice. More notation.

Nothing more than that.

He slept.