User blog comment:Ygrain/EDI and the geth: their purpose in the story?/@comment-4953843-20120723181742

warning: TL;DR. EDI and the geth were supposed to be -the- counterexamples to the catalyst's assertions (and the only things preventing me from destroy).

why in the billions of years of reaper existence had the catalyst not encountered benevolent ai, you might wonder?

consider the impartial experimenter analogy, with the catalyst as the observer, the galaxy as its testing field, the reapers its testing method, and the organo-synthetic conflict as its hypothesis.

billions of years of repetition always producing the same results might have led it to accept its assertions as fact. a skynet/omnius/asuran collective every 50,000 years. and as a machine, if it ever did encounter cases of benevolent ai, it would have been statistically insignificant compared to the other cases it uncovered. for all we know the reapers may have had harvested cycles where only synthetics were left.

for the first few cycles it might have worked.

except that it consistently applied controlled experiment rules to something that has far too many variables to be totally controlled (galactic evolution) and in doing so failed/refuse to notice its own gradual contamination of the samples (zha'til - means to an end, and again, corruption timeline unclear). observer effect. compounded errors over the cycles = crucible blueprints.

enter the variables of its last cycle. mutated keepers. sovereign taking too long to assess the situation. human genetic diversity (which, apparently, is a big thing). deployment of the crucible. and even an antithesis emerged: organo-synthetic cooperation. the variables changed radically, the catalyst chose to continue anyway (it still had technological/numerical superiority with which it hoped to rein the experiment back in line - the purging phase, at least). but when it finally revealed itself it even admits that its solution was no longer viable - that's because of the drastically altered variables. if you refuse, you're just validating the reapers' position until the next cycle does it right.

point is, none of the previous cycles that -may- have had benevolent ai in them mustered enough counterevidence to prove catalyst wrong: no cases of organo-synthetic cooperation emerged powerful enough to challenge starkid's ideas on his throne on the citadel. until a certain shepard appeared with his/her AI ship (with optional geth if s/he didn't destroy them - if i destroyed the geth i would have no problem choosing destroying reapers as well).

again, billions of years of the same experimental results + sloppy execution and cleanup per cycle + humans are special = the most unbalanced cycle the catalyst ever experienced.