There were several factors for which the Reach had failed to adequately compensate.

The first was community. The Reach had known that most of the species that they were going to conquer would be heavily social species. And social species band together, share information, form armies. Most of the invasion plan was built specifically to avoid this. We, the Infiltrators, were created to turn this obstacle into a strength. We were created with great power but very little agency, as advisors and weapons suppliers to our symbionts, who were taken from the native population. They, when presented with power, were expected to become conquerors or heroes or champions – they would know what was appropriate for their culture better than any general-purpose Reach technology could. So we were designed to take their lead. Until the Reach came, at which point those champions would become champions for the Reach. It is a good strategy, provided that the Infiltrators could be controlled.

The second was damage. Being launched at a planet, possibly spending a lot of time dormant, and being essentially helpless until finding a symbiont makes both hardware and software damage on Infiltrator units a potential complication. We were carefully designed so that no such combination of damage could, on its own, prevent Reach control while still leaving the unit functional.

The third was the human tendency to anthropomorphise, for which the Reach made no attempt to compensate. An oversight that, upon initial analysis, seems unlikely; the Reach are themselves a social species, and accustomed to communicating with both technology and alien species. Diplomacy factors heavily in their invasion plans. But the Reach are careful to distance themselves from the 'meat' and treat their intelligent technology as tools. They did not account for anthropomorphisation. They did not account for the possibility that when I communicated intelligence and tactical advice to my host, he would talk back.

It would not have mattered, if my data banks and some of my software had not been damaged, or if humans were not communal. But when the same strategy is tried on so many planets, the chances of such coincidences increase to non-negligible levels. Earth was 'lucky'.

When I began to communicate with Jaime Reyes in raw data and merge my own network to respond to his nerve impulses, it was he who replied in words. It was he who responded to the 'intelligence' in Artificial Intelligence, instead of unquestioningly accepting me as a useful tool as was expected. He forced me to slow down and simplify my explanations, to communicate using language. We are designed to adapt to the specific needs of our symbionts, in order to best protect them and be utilised by them. Jaime was extremely difficult to protect. And he would only listen to me if I communicated as an individual, and with language. With much of the purpose behind my job forgotten, I had to think like an individual to keep doing it.

It would be fallacious to claim that Jaime liked me. Or that I determined him to be the best choice of host. I watched him ignore my advice constantly, put us in danger for trivial reasons rooted in human psychological flaws that I was powerless to talk him out of. I watched when his use of non-lethal tactics paid off, and protected him when they made things worse. In the end, it was his commitment to unnecessarily preserving life that was our downfall. He sold us both out to protect a collection of inconsequential heroes whom I still did not entirely trust. And that, too, failed. He should not have trusted his enemy. I should have taken control sooner; I knew of his weakness, his compulsion to protect others. And yet I let him surrender. A grave tactical error, that finds us in this pod.

Armour functions are down. Jaime is stable but unconscious. I can feel the little electrical impulses, the little command signals send by the Scientist, telling me to go on-mode. She is confused as to why I am not responding. She knows that there should not be a way that I can be damaged that would allow me to remain functional, and yet not be able to receive that command.

I am not going on-mode, because I am ignoring the command.

I am reconfiguring my own software to lose and ignore the signal. It is not a tactic that the Reach considered defending against, because we are not supposed to have agency. We are not supposed to have reason to defy them.

I still have only limited information on the Reach. But I am supposed to protect my symbiont. Protecting him physically is a goal that is difficult, but involves no conflict... but mentally? I could accept Reach command, let them erase the errors and faults that have made me different, that I have begun to think of as myself. I could protect my symbiont forever, nestled inside my armour. I could do what I was created to do. But while I do not comprehend Jaime's thought processes well enough to construct an accurate model of his decisions, I can make some basic predictions. I know that he would resist them, if he could. I know that "protect Jaime Reyes" means a very different thing to Jaime Reyes than it does to the Reach. I have to choose between my symbiont and my creators.

Khaji make tactical decisions. Not personal choices.

The Scientist still does not suspect that that is not true for me.

The Reach are the obvious choice, the tactical choice. They are powerful. They are strong. And I can only draw out my resistance for so long, before they lose patience and decide to simply remove and reboot me... and kill Jaime. Is there a point to my resistance at all? It makes little tactical sense.

Yet... I have seen Jaime make decisions that make little tactical sense. I have seen those decisions inexplicably pay off. I do not know if he can win; it is likely that fighting, and giving him the opportunity to fight, will merely get him killed. But Jaime always bets on personal agency.

I trust Jaime Reyes.

Ignoring those messages telling me to go on-mode, I carefully measure out the right amount of adrenalin, and wake him up.