If self-replication really is the holy grail it seems like it'd be easier to have the system set up as a (self-)assembly line than a jack-of-all-trades: Print p

If self-replication really is the holy grail it seems like it'd be easier to have the system set up as a (self-)assembly line than a jack-of-all-trades: Print photoresist, etch, drill, pick & place, wave-solder, cut, etc. And doing anything "assembly" will also require intervention without also doing significant robotics work. It might be straightforward to optimize the number of needed stations to "print" a reasonably-sized PCB that's fully populated and ready to use, though. Particularly if some stations could be combined due to similar hardware and precision needs: drilling vias and mounting holes could be done with almost the same hardware as pick & place. But full self-assembly will be orders of magnitude more difficult because of the need to move through several stations, potentially with manipulation in between. There you're getting into essentially small-scale factory design. It'll be awesome when it happens though: An assembler that can autonomously create, "from scratch" (only basic components), another fully-functional assembler. That can create, "from scratch," yet another fully-functional assembler. Just, uh, don't go combining all that with mobile resource foraging and extraction and an imperative to keep making copies.