Also called: RPU
After the Fall of the General Galactic, when people started to warm to the idea of some kinds of AI, they created Reality Processing Units. Originally they were only for very large and dedicated simulations and took up around 20,000 cubic meters of space. By the start of Intergalactic Expansion they had shrunk to the size of a single chip and could be used to create Reality Boards.
Reality Processing Units were loaded with a Contained AI Environment. The unit's physical connections provided a way to read and write data within the Contained AI Environment. Before the scale reduction allowed for the creation of Reality Boards, Reality Processing Units had broad purpose Contained AI Environments that had to be manually programmed. Depending on the nature of the needed calculations creating this kind of system could take teams of people years to develop. Once the RPUs were small enough though it was more practical to have several of them work in tandem, which is where Reality Boards came into being. There was an initial feedback loop where each RPU was able to be loaded with more generic and smaller Contained AI Environments, so the hardware needs of an individual RPU simplified, allowing them to get smaller, so more could fit on a Reality Board.
When a Contained AI Environment would give bad calculations, the RPU was transferred out of the main system and into a special CAIE Interface. From there one or more CAIE Troubleshooters would work on the Contained AI Environment and determine if it needed to be patched or if the whole RPU needed reimaging. Eventually the set of Contained AI Environments was generic enough to do away with this step for end users, who could just reimage any time an RPU was giving bad calculations. There was always need for CAIE Troubleshooters within the companies developing boilerplate or highly custom Contained AI Environments.