Finding the frequency wasn’t a fluke so much as the dedication of crackpots. The efforts of ghost hunters and paranormal investigators whose sustained pseudo-science crackpottery was able to capture enough samples of the ghost frequency to eventually warrant attention of the scientific community.

It was years before investigations looped back around to give some measure of validation to the paranormal community. The localized frequencies weren’t coming from space, weren’t bleed off from existing EM emitting sources, and had all the hallmarks of intentionality.

It was longer still before it was determined they weren’t attempts at external communication, the sources weren’t trying to talk to each other or us, but were bleed over from unshielded, internal processes.

Eventually, enough samples were gathered and examined by programmers for it to be determined the signals were code. The capture of faint, unshielded systems running through their internal programs. The bleed of an ephemeral operating system executing.

The syntax wasn’t complex when it was finally deciphered. When enough keywords and types were identified to start making sense of the instruction sets. It was a low level language, some type of assembly code written to a strange and unknown processor whose nature could only be inferred from what was understood about the code running on it.

It was an adjusted binary system which had to take into account overlapping operating environments. There were only a handful of functions focused on basic motion through the two environments, a limited number of actions, a limited number of sounds, and a limited number of inputs.

Once enough of the code had been collected and deciphered to allow for basic writing of it and equipment had been designed to broadcast as well as receive the frequencies, work began on hacking into system.

Success was limited at first. Though the code base was simple, exploits were difficult to find. The language seemed not be designed for extending, and once an instance of the code was running, once an individual ghost was created, it seemed to be designed to run in a loop with a limited number of possible branches. It was hardened against code injection by virtue of the simplicity of the language and the standalone, repetitive nature of the ghosts.

The exploit initially discovered had to do with the program’s termination parameters. When the ghost received a certain set of inputs which varied from ghost to ghost, but which fell into a fairly small group of types, the program had the ability to terminate itself and for all intents and purposes the ghost disappeared.

At first this allowed only for early termination. If enough of a ghost’s code could be gathered to identify what termination input it was waiting for, the input could be generated and broadcast, forcing the ghost to shut down. The public called this ‘ghost busting’ and the paranormal community latched onto the process to commercialize it. Businesses sprang up, reality TV shows, a whole industry.

The scientists, though, saw this as only a baby step, and not a good one, as in forcing the ghost to terminate, they also lost their connection and ability to collect more data.

Eventually, through trial and error, it was found that new types of input, outside the parameters of the original design, but incorporating some of the stimulus being looked for by the OS could cause the ghost to glitch, and from that they were able to refine the process until a pattern was discovered which could force the ghost to incorporate new instructions as valid and executable.

At this point, the ghosts became hackable.

This level of work was classified and owned by the government, of course, which also meant it ended up being weaponized.

A manual was created defining the input and output parameters of the OS, the set of functions the OS was designed to accept, the set of known non-native functions which could be reliably injected, and a syntax to be used in designing new non-functions.

The most obvious use of hacked ghosts were spying and messaging. A basic AV capture routine could be passed into a ghost, then instructions for the ghost to go to specific coordinates, run the capture, then report to a retrieval station where it would give up the captured data to be reconstituted for use.

Limitations of the frequency made it difficult to communicate with the ghosts over distances, so initially all the requirements for an operation had to be coded at the beginning, hoping the ghost wouldn’t run into unexpected input and conditions. Later, boosters which could successfully increase the range of signaling were developed, but it was never a situation where a station anywhere in the world could control a given ghost anywhere else.