I was enjoying the ADHD hyper focus writing, kind of following along, then:

  > If we figure that the foil can meaningfully change direction every 20 μm, then we might think of an aluminum-foil machine as being made of “moving parts” on the order of 1000 μm² (50 μm × 20 μm), 1000 “parts” per square millimeter of foil; a roll of kitchen aluminum foil is enough to fabricate some 4 billion “parts”. A bootstrapping compiler might require 100 000 parts and thus a square centimeter of aluminum foil, cut and folded around into a shape a couple of millimeters in diameter. If it were doing only one thing at a time, and needed 10 seconds to construct/assemble each moving part, it would take about 12 days to recompile itself. This is probably adequately fast, barely, but probably not adequately robust against errors. It would probably be better to design it to have more parts and do many things at once, enabling it to be faster and correct errors.

I do not currently have a sketch of a matter compiler, just a rough estimate that most of its parts count would probably be RAM, and that it probably needs several kilobytes of RAM.

I've read a lot of his other writings so that context might be informing my reading here but it sounds like he's pretty straightforwardly discussing the potential of aluminum foil as a uniform-feedstock-slash-construction-material for a hypothetical self-reproducing microfabricator.

It reads like a rambling of a mad scientist, but I think simply adding pictures would make this feel absolutely normal (in the crowd of other blogs featured on hn)