He noticed the railing posts first.
Not all of them. Specific ones, in specific stretches. The standard spacing between posts was roughly two metres. He had walked past enough of them that the interval had settled somewhere below conscious attention, the same way the bridge hum had settled. Two metres was normal. Two metres was the bridge at rest.
What he began noticing was that in certain sections the spacing tightened. Not dramatically. A few centimetres less between posts, sometimes half a metre less, the change too gradual to catch in any single step but measurable across a ten-metre run if you were paying attention. He had not been paying attention to this specifically. He had been paying attention to everything simultaneously, which was a different thing, and patterns tended to surface from that kind of attention without being sought.
He stopped where he was and looked back along the last hundred metres.
Three sections with tighter post spacing. He ran the events of the last few days against them. The hollow resonance before the combo trap: tighter posts. The floor drop before he acquired the vine arm: he had not checked then, did not know for certain. But the section with the needle chest had tighter posts too, he was fairly sure.
He looked at the section ahead. Normal spacing. He kept walking.
Two hundred metres later, tighter posts. He slowed without deciding to, the change registering before he had processed it as information. He pressed each plank with his foot. Fourth plank in: the extra flex. He shifted around it and moved to the edge of the bridge and continued along the railing until the post spacing normalised.
No trap fired. He had not triggered it. He had simply walked around it because the section had told him it was there before the plank confirmed it.
He stood at the far end of the section and thought about that for a moment.
The wood grain was the second thing.
Certain planks in trap sections had a grain that ran differently from the surrounding boards. Not obviously. Not the kind of thing you would see from standing height unless you were already looking for it. But when he crouched and looked at the plank surface at low angle, the grain near release mechanisms tended to run perpendicular to the plank's length rather than along it. The boards had been cut differently. Probably from different stock, installed for a purpose rather than as general flooring.
He did not test this hypothesis by triggering anything. He did not need to. The pattern was useful as a pre-confirmation, not as a primary tell. The flex test was still the primary tell. But now he had a reason to crouch and look before he pressed, which meant he was reading two signals instead of one, and two signals that agreed with each other were worth considerably more than either alone.
He also noticed, cataloguing the post spacing pattern, that it was not perfectly reliable. In one section the posts were tight and the planks passed the flex test and the grain ran straight and nothing fired when he crossed it. The bridge offered no explanation. He noted the exception and adjusted his model: tight posts were a warning, not a guarantee. Worth slowing for. Not worth stopping at indefinitely.
Better, not certain. The gap between those two things was still where the bridge lived.
The first mark went on the left wall about three hours into the day's travel.
He had pulled one of the arrows from the combo trap railing earlier, working it loose from the post with his right hand while the scooter idled beside him. The shaft was intact: dark wood, straight, the fletching intact. He had looked at it for a moment, turned it over, found the tip still sharp. He pocketed the shaft without examining what the impulse was. It had been in a marketplace listing once. He had no immediate use for it himself.
Now he used the tip to scratch a horizontal line into the left wall at shoulder height. One line. He stepped back and looked at it. A mark on the bridge that had not been there before he put it there. He had passed approximately three hundred metres by his rough estimate, accounting for the scooter runs and the sections where he had walked more slowly.
He was aware that the estimate was imprecise. He was aware that if he never needed to look back the marks would serve no purpose. He was aware that the bridge was long enough that a few scratches were not going to constitute a useful map.
He made the mark anyway. Then started the scooter and kept moving.
He made the next mark fifty metres later, at a place where the walls came close enough to reach from the bridge. A second line beside the first. He kept the system simple: lines for rough distance, a cross through a line for a trap section cleared. At the next section he cleared he scratched a cross. The cross looked like the inside of a trap chest lid and he noticed that and moved on.
The scooter changed how the marks felt.
Walking, he had measured distance in time. An hour of walking was a unit. On the scooter, the same distance passed in minutes, and the marks on the wall came up faster than he had expected, each one confirming the distance covered, each one showing him that he was moving through the bridge at a rate his legs alone could not have managed.
He stopped at a section where the walls were wide and the post spacing was regular and the wood grain was straight and nothing about the section read as preparation for anything. He idled the scooter and looked ahead. The bridge ran out to the familiar grey limit. He turned the throttle and rode.
He rode for six minutes before the walls tightened and the posts closed and he braked and stepped off. His ribs registered the deceleration. He stood beside the scooter in the silence that followed the engine's idle and estimated the distance covered. Close to a kilometre. Perhaps more. What had been a day's careful walking was behind him.
He pressed the nearest plank. No extra flex. He checked the grain. Straight. He pressed the next plank. Straight. He got back on and rode the section at half the previous speed, foot testing the platform edge at intervals, and nothing triggered and he came out the other side with the post spacing normal and the walls returning to their wider position.
Wrong prediction, or incomplete one. The posts had tightened but the section had been clean. He noted it the same way he noted the earlier exception: the warning was probabilistic. He revised his confidence in the post-spacing tell slightly downward and kept moving.
He found a chest near the end of the day's travel, a single horizontal mark. Water inside, which he needed, and a second block of dried food, and a coil of thin metal wire he did not immediately have a use for. He drank half the water on the spot. He put the wire in the scooter's storage slot next to the near-empty fuel canister. He did not know what wire was worth on the marketplace or whether anyone needed it. It weighed nothing. He had the space.
He made his last mark of the day at the chest's location, a third line on the left wall. He stepped back and looked at the three marks in the stone. The first was three hundred metres back by his reckoning. If the estimate was right, each line was roughly a hundred and fifty metres apart.
The marks did not tell him where the end of the bridge was. They told him only where he had been, which was a different kind of information and not the kind he most needed. But he kept making them because the act of making them was different from just walking. It was the smallest possible act of claim over a place that had given him nothing but difficulty, and it cost nothing, and he was going to keep doing it.
He switched the scooter off and sat against the left wall beside his newest scratch and ate the food block and drank the rest of the water he had allowed himself and watched the bridge for a while.
He was no longer the person who had stood at the entrance and hesitated before the first step. He had not stopped being afraid of what the bridge might do next. But the fear was organised now, filed alongside everything else, subordinate to the reading and the moving and the marking of walls. That was the shift, and it was complete enough that he could see it clearly, and he did not think about it longer than that.
He started the scooter, pointed it forward, and rode.