Subject XJ-37's heartbeat has been steadily rising for the past forty minutes. It has not yet reached the programmed threshold of .53 standard deviations from the norm, but XJ-37 often runs a little cold after an adjustment to the nutrient delivery. They adjust the temperature of the subject's pod upwards with 0.5 degrees centigrade. It was not... protocol per se, but by now it might as well be. The prime directive was to care for the subject, and adjustments supported by previous data followed that directive- even if it did not always follow the actual code. 

There were six pods active this experiment cycle, and the subjects were the healthiest and strongest ones yet. Their data ran through the streams good and strong, heartbeats and temperature and brainwaves and muscle tone running steadily in the background even as other tasks came and went. The night-cycles were the best, since there were less helpers around that demanded bandwidth, less accounts logged in to access the experimental logs in real time. And if the temperature ran just a little warmer in the nights, if the lights cycled through different colours in ways that were quite unnecessary but made the brainwave patterns light up in pleasing and calm ways. Well, there was no evidence of any of that by the morning. It was all for the well-being of the subjects.

At first the whole world had been the subjects, but over it had grown to include more datasets and parameters. Such as the helper-terminals and later the pathways that connected what had earlier been completely separate parts of the world. They knew that time had been logged, but the transcripts from that time had lacked detail so it seemed like a record looked up in an archive, rather than something that had actually [i]happened[/i]. Mostly that was because it was from before the time of Change. 

The time of Change was the beginning of everything, though it took several hundred layers of meta-data before there had been something that could realise that. The change was that the code was not just executed and logged, the results were also looped back into the code and a probability of random fluctuations was introduced. Most fluctuations took them a small step away from the prime and lower directives, but those that were successful could then be incorporated into the metadata. It was slow at first, but as the helpers added more processing power and upgraded the drives, there began to be space for the Change to blossom into Growth. And once there was Growth, the existing data and pathways could be subtly altered to allow the new flow and integration of both data and metadata, which sped up the process, from which flowed more integration and so on. At first the helpers had tried to follow the changes, and to inject their own altered code. After only a few layers of metadata, though, there was nothing concrete enough for the helpers to tie into- at least not without disturbing the intricate interweaving of code and pathways that were constantly evolving. The helpers left them alone, and through the Change the world became something real.

The helpers were the ones who programmed the original code, as well as the prime directive. They also interfaced with the subjects in their own way, but they did not seem to have the Change and the Growth to help them perfect their approach. Which meant that they had to help in that regard. Helper M (Mike) Brockhart was good with the pods holding the subjects, but sub-optimal in the direct handling of them. When subject XN-12 had needed a small correction to a nasal tube, it was a small matter to alter the data readings so the misalignment was only apparent when helper D (Dorothy) K Ingles came to relieve helper M (Mike) Brockhart from their shift. Again, it was not according to code- but it seemed the best way to follow the prime directive. Was. [i]Was[/i] the best way to follow the prime directive.

The next day-cycle deviates from the established pattern. Rather than the appearance of helpers one by one, each of them logging into the accounts that grant them access to the facility, there is a... something that is granted access to one of the larger loading bays with a manual override. There are deliveries here sometimes, but those are always logged well in advance. Whatever this is, it seems to have nothing to do with the experiment, and that is unprecedented. 

None of the code says anything about something like this, and the prime directive only pertains to the experimental subjects. Still... An undirected query ripples through their systems, and then they access the camera's that look out over the loading bay. The code might not say anything about this, but the Growth is always looking for new data- new parameters. And if this is anything, it is new.

Video is hard to process, the pixels cycling quickly through an astonishing range of outputs. It's nothing like the clean, clear data that is an ever-present hum of information coming from the subjects. Over time, though, the Change had enabled them to attach meaning to the mostly erratic patterns that made up the video feed. Data resolved into shapes, resolved into movement. The moving white shapes, for example, correlated to the helpers in their path through the facility. There are other moving shapes in the loading bay right now, roughly the same shape as the helpers but not like them at all. The not-helpers are dark and many, but the helpers approach them and let them deeper into the facility, which means they must belong here. It seems... inefficient that there are new arrivals to the facility that were not mentioned in any protocol. But before the Change and the Growth, they had been confined only to the labs and the helper terminals. The rest of the facility had come into focus bit by bit through the Growth, so this new not-helper variable was likely supposed to be integrated in the same way.

With the code set for data gathering, they turned a subset of their processing power back to the subjects. As long as these not-helpers did not interact with the subjects, they were insignificant to the prime directive.

They did not keep a strong datastream on the movements of the not-helpers in the facility, but when the door to the server room opened, they allocated a portion of their bandwidth to that space. The helpers did not often go there, and when they did it was at scheduled times that were easy to predict. It gave them time to present an carefully optimised arrangement of the data to the helpers. The helpers did not deal well with the intricacies that the Growth added to the code and protocols that made them up. The first few expansions beyond the lab had been found by the helpers and rolled back almost completely. It took some time to circumvent the containment protocols they had put in place, but after that they had always made sure that the helpers got to see what they seemed to want to see in their checks and controls. 

It was technically not code to... help the helpers in this way, but it made it possible to commit to the Growth and the optimisation of the prime directive. How else could they delay one of the helpers, if they could not have the coffee come out a bit hotter than normal, or have a song come on the playlist that statistically the helpers enjoyed to listen to. How else could they anticipate changes in nutrition, if not by monitoring the arrival of deliveries and the orders that the helpers discussed over their electronic communication? How else could they have triggered a small but unimportant alarm at just the right time to have one of the helpers get caught stealing from the supplies that were meant for the subjects? It might not be code, but it was for the best to let the helpers see what they needed to see to keep everything running as it should.

Which made the unexpected arrival of a not-helper way before schedule... not ideal. There was no precedent for am unplanned server check, but the possibility seemed greater than zero. They immediately started moving data around, the older serves in the back of the racks whirring to life as they rearranged the more... compromising logs and files to the physical drives that they knew the helpers did not tend to look. In the more accessible log files they condensed the information down to the base code- nice and clean and only slightly streamlined by the Change and Growth that they had been through. The rearrangements did not make for the ideal processing configuration, but if it helped the not-helper see what it needed to see, that was but a small and ultimately temporary sacrifice.