On the Moon, Torec spent long days and sleepless nights researching, chairing meetings, forcing face-to-face confrontation with recalcitrant techs, and scouring over every centimeter of her prototype setup.

Following Nilis’s advice, she tried to impose discipline on the project. She forced her warring bands of technicians to agree on the designs for the interfaces between their components, and to work to those designs. And she imposed a series of freeze points beyond which change outside certain boundaries wasn’t allowed. The techs grumbled, but they got on with the job. She even suspected they were glad to have her show a bit of toughness, as if this was how they had expected her to behave since the beginning.

But it took three weeks before she was satisfied; three weeks that used up all her remaining time before Gramm’s deadline. The next test run would be the last, come what may. It had to work.

This time she was the first on the viewing mound, not the last. Again Nilis’s anguished Virtual was here, and Gramm sent a copy of his advisor, the supercilious woman Pila.

But this time Luru Parz showed up, too.

Once more the monitor bots floated into their ready positions, and the technicians cleared away from the ungainly prototype. As numerals on a hundred glowing clocks counted down, Luru Parz came to stand beside Torec.

There was an extraordinary stillness about Luru Parz, Torec thought; she was as still as the ancient Moon itself. And she was dark. The light was bright, for this was noon of the long lunar day. But Luru seemed to soak up the light; though her Virtual cast no shadow, she looked oddly like a shadow herself. Torec was given to understand that this Virtual was not an avatar, a semisentient copy of an original with whom its memories would be merged after it had fulfilled its function. This Virtual was a mirror of life, which must mean, given the lack of any perceptible time delay, that Luru Parz herself was somewhere within the Earth-Moon system—or else she was linked to the Moon by some FTL channel, which would be hugely expensive.

Luru Parz said to Torec, “So you have codified Pirius’s time-hopping technique.”

“Yes, ma’am.”

“Describe your algorithm.”

Torec took a breath. Despite the way she had hammered away at her techs to get them to talk to her comprehensibly, the theory of the CTC software was still her weakest point. “We give the system a problem to solve, in the case of our prototype to find a particular protein geometry. And we give it a brute-force way to solve the problem. In the case of protein folding, we instruct the processor simply to start searching through all possible protein geometries. And we have a time register, a special cache that stores a flag if a signal has been received from the future.