When White had first unveiled Pipistrelle to Holland and Knox in its full miniature glory, they’d immediately seen its potential. Even the most secure rooms, buried deep in embassies and wrapped in layers of soundproofing, would become vulnerable to MI5-eavesdropping. Knox also couldn’t believe how small and deceptively innocent it looked. Somehow White had managed to fit Pipistrelle’s battery, aerial and transmitter in a simple square metal box half the size of Knox’s thumb.

‘How did you get it all in there?’ Knox asked White as he nudged the tiny prototype across his palm with his fingertip.

‘A lot of hard work,’ White replied.

Pipistrelle was Britain’s greatest intelligence weapon since Bletchley Park cracked the Enigma code, and was years beyond Venona, the joint project between the UK, America, and Australia that was still decrypting decades-old Soviet cables. For the last three years it had provided MI5 with invaluable intelligence and remained completely undetected. However, six months ago White had revealed the next step in Pipistrelle’s evolution and it had driven a wedge between him and Knox.

‘Our problem,’ White had said when he presented his new proposal to Holland and Knox, ‘is processing power. We need to be able to work on everything in one place, as soon as we get it.’

‘You mean using a computer?’ Knox asked.

‘I’ve read the budget request,’ Holland answered. ‘He means more than that.’

‘A supercomputer,’ White replied. ‘The University of Manchester’s new Atlas machine is more powerful than anything IBM or CDC are working on in America, and far ahead of what we think the Russians are up to.’

The Atlas was built on a solid-state germanium transistor infrastructure, using 128 high speed index register and 680 kilobytes of memory, which was roughly the same amount of storage and processing power as every other computer in Britain put together.

Knox understood the value of technology—Pipistrelle had proven itself time and again—but he wasn’t ready to hand everything over to it. He believed there were still some things machines just couldn’t do.

‘How many analysts could we hire for what this will cost?’ he asked.

‘Atlas can analyse much more information than a person, and in a fraction of the time,’ White replied.

‘But it can’t make decisions, or draw conclusions.’

‘Nor is it burdened by habit or narrow viewpoints,’ White said, defensively. ‘It uses every variable to make the most likely predictions.’

‘Unfortunately,’ Knox countered, ‘our enemies don’t tend to do the most likely things.’

Holland cut their argument short, approving White’s request on the proviso that flesh-and-bone analysts would continue working on Pipistrelle intelligence and check everything that was run through Atlas.