It was in this aspect of the Delilah that his cryptanalytic experience came into play.
The work they had done so far constituted the mechanism for 'adding on' to speech.
The crucial question of what to add was the one on which he had spent much of his time since 1938.
In this he could act as the mathematical Cambridge and Bletchley figure, rather than someone who had joined in, somewhat awkwardly and embarrassingly, with the expanding world of electronic engineering.
Although he could not say, nor even hint, the task amounted to creating something like the Fish key generator.
It had to be deterministic, for otherwise it could not be produced identically at two independent ends.
But it had to be sufficiently devoid of pattern or repetition, to be as secure as something truly 'random', such as electronic noise.
Any kind of mechanism would, inevitably, have some pattern to it; the job was to make sure that it was one that the enemy cryptanalyst could not possibly detect.
So in doing this for the Delilah, he was finally scoring off the half-hearted efforts of German cryptography.
In fact, he was doing very much better, for the Delilah key would have to be supplied in sequences of hundreds of thousands of numbers.
It was like enciphering not telegraph messages, but War and Peace.