ａｕｔｏｐｏｉｅｔｉｃ　ｒｅｔｕｒｎ　ｏｆ　ｍｅｃｈａｎｉｃａｌ　ａｓｓｅｍｂｌｙ：

return

imagine a robot arm.
the robot arm has several joints, giving it the semblance of an 'arm' instead of a 'numerically controlled pincer', which terminate in a tool not unlike a pair of pliers.
in ideogram: 

    'D--o--o--C'

where a 'D' is a hinged base, a '-o-' a servo-controlled joint, and a 'C' a grasper.
what is noteworth of such devices, as accumulate in the halls of engineering colleges and workshops of more luxurious industrial parks, is the system of tools used to assemble them, and the system of tools these 'robot arms' may themselves use.
setting aside those *boring* and *incomplete* critiques such as 'what if the arm cannot assemble *itself* before it is yet built? huh? huh?', a more interesting critique appears before us: what sort of robot arm can turn the screws which join together its own rigid members to its base? which join its servomotors to its hinges and swivels? which join its plier-fingers to the flexible, controllable arm-wrist? 

    CCC        CC
    CCCCC     CCCCC
    CCCCCC    CCCCC
    CCCCCC   CCCCCC
    CCCCCC   CCCCCC
    CCCCCCCCCCCCCC
    CCCCCCCCCCCCCC

it turns out you don't see many wrists on 'robot arms' which can robustly seat a screwdriver's tip in any of the screw-holes which constitute their own bodies. this problem itself excludes all sorts of other interesting questions, like how one learns to balance a screwhead upon the tip of a screwdriver while maneuvering driver-screw to screw-hole. slotting screws into their right positions seems to require a mixture of highly flexible low-range movements, after various coarse and imprecise situating-movements across long distances. a robot arm which can only poise screws on screwdrivers, making the screws *ready* for use in the assembly of a robot arm is apparently not the sort of arm which can repeat its own assembly.
if we pull back for a moment, can we make this question a little bit simpler for ourselves? it, after all, is not very important why a screw is hard to poise on the tip of a screwdriver, or why it is hard to align a screw to a bore-hole in a slat or metal cylinder. maybe a better question we can ask is this:
if a mechanical tool is made from 3 assembly actions on 6 substitutable parts, a tool which completes at least all 3 assembly actions on at least all 6 substitutable parts is self-kind-assembling. (for those intoxicated by string substitution rules, a 'n' and 'k' might be neatly slotted in place over those '3's and '6's.) this is a dull and straightforward definition: it speaks not to exciting mysteries of 'where the tool come from?', 'where the tool go?', or even 'cotton eye joe???'. but it sure is better than counting 'degrees of freedom' or 'mega-milli-macro-meta-mesa newtons per pound-foot of joule^3/log(summation(choices))'. i sure don't see a degrees of freedom counter or a minimum (or is it maximum?) shannon-bits rating on the handle of any screwdriver i've ever used!
what this tells us, i think, is that the sort of robot arm which can assemble its self-kind is one which is deliberately tested and evaluated against mechancial assembly tasks. that the robot arms poised in oh-so-many engineering colleges are graded by joint-count or newtons or coulombs or, worst of all, years of calculus prerequisites (😂) has as little inhibitory hold over the assembly of robot-arms as the robot-arms in question have hold over screwdrivers.
it appears that some other kind of 'greater robot' will definitely pay tens of thousands of dollars per year of tuition to screw together exactly 3 servoes in series. if and only if the servoes are inside a college campus, not a kitchen table at home. if and only if the pincer-arm is pinned together *after*, not *before*, a series of weed-out classes on electrical field equations.

(wait, did you think this text was about using mechanical assembly of parts to recover and restore the necessary conditions to mechanically assemble more parts? quite the opposite! why then would this text ever mention pincer-ended pedagogical robots built from screws? we use robotic *welders* for commodities that matter. are you timelost from 1987? go look at a unibody macbook or something!)

one might note, with some sorrow, that the absurd precision and reliability of benchtop laser cutters in the year 2025 comes from the *laser-cutter*, a fantastical and wonderful tool which belches noxious odors and literally poisonous fumes in exchange for a cutting-action which may be repeated for dozens of meters of cutting-work without losing steering-calibration or cutting-strength. the precision and power of the laser-computerized-numerical-control robot is carried entirely by its laser, it appears, not the computerization nor numerical control. could it be that all of those human generations fastidiously building pedagogical robot arms returned us no new numerical control, no computerization of control, and no computerized-numerical-control of mechanical assembly?


...


i mean yeah. yeah. of course it didn't?