BTW I reported the input/output problem to WUBE, and they fixed it for .45. Once .45 is pushed in stable, I'll revert the behavior to a bidirectional fluidbox
I saw that in the changelog for .45 and your post on the forums as I like to look at the user reports that lead to the bug fix. While I don't agree with his coping and seething, I do think the new system does warrant at least discussion. I do not see the purpose to having the fiber optic cables (HIGH SPEED INTERNET ACCESS!) be bi-directional. I used to do networking infrastructure for OC-192 backbones for colleges and businesses, and they ALWAYS had 2 cables, one for downstream, one for upstream at the bare minimum. Why? Because fiber optic cable (HIGH SPEED INTERNET ACCESS!) works by sending 1 (light on) or 0 (light off). I don't know if it's theoretically possible to have bi-directional fiber optic cable (HIGH SPEED INTERNET ACCESS!) without establishing some kind of communication schedule ahead of time.
Anyway, I've since learnt that fiber optics has gotten exponentially more complicated, however, it's just because they added more bonded pairs, not exactly the best solution.
However, you could argue that the AI you're building is so vastly intelligent it knows how to perfectly demultiplex the signal by something something AI something something science.