Seeing and testing are different jobs, done in sequence

He saw things that fit. That was the whole of what he did, and she had spent three weeks treating it as a flaw in his method because it wasn't hers — because a claim that arrived fully formed, stitched together out of a dozen disciplines and two centuries and a species' worth of half-remembered myth, looked to her like something built to survive scrutiny rather than something that had earned the right to. But he wasn't hiding from her gauntlet. He simply wasn't built to run it himself. He saw the shape a thing wanted to be, and he built toward that shape with whatever material was closest to hand, real or half-real, checkable or not, because the shape was the part he trusted, and he trusted it the way she trusted a test result — completely, without needing anyone else's permission to.

What she was, was the thing that came after him. The part of the process that took a shape somebody else had seen and asked it, honestly, patiently, without mercy, whether it could survive being disbelieved. He built. She tested. And a discovery that had never once been tested — however beautifully built, however many real anchors it was stitched from — was not yet a discovery. It was only an excellent guess, dressed in enough true things to be mistaken for one. What made it real, what made it something the next room full of skeptical people could actually stand on, was not the seeing. It was the surviving.

He had seen it first. She had made it trustworthy. Neither of those was the lesser art, and she understood, finally, that she had spent three weeks quietly competing with a man who had never once been playing the same game she was — not because he'd rigged it, but because there had only ever been one game on this ship that mattered in the end, and it had needed both of them to finish it, and only one of them had ever fully understood that.