Note: I'm posting this partly because I recognize the thoughts therein are rough and have gigantic logical holes in them, and welcome people to blow cannonballs through them to force me to make my thinking on this topic more rigorous. Also, I'm really bad at fabrication.

An email discussion started around the multimachine very late last night, and a friend brought up (half-seriously) the question "so what would it take to fully digitize it?"

My reply:

You can CNC-ify a lot of things. To do it with this particular machine would probably involve a somewhat awkward series of contraptions to move (with great precision) either Very Large Motors (the multimachine) or Very Large Parts. But I'm guessing that's why you asked the question with a grin in the first place, and it's the reason I'm going to start on my own monologue here in reply.

Disclaimer: I'm a coder and electrical engineer, and I love digital fabrication and the idea (and practice) of treating physical stuff as information to be manipulated. I think it's a tremendously powerful one. I think that more work in this area can and will happen, and that the resulting developments in personal fabrication will revolutionize the world and make great changes and be an enabling factor and all that stuff.


There are some things that analog fabrication can do that digital can't, at least not at the moment (and probably for a while). Sloppy, inaccurate, non-mass-producible, human-time-required analog fabrication. Those seeming weaknesses are also, under the right circumstances, strengths.

Sure, we can make welding robots (heck, car factories have been using them for years). But when you're monkeying around an old steel bridge in Jakarta that needs a couple handrails stuck on, it's far faster, cheaper, more efficient, and probably just better design to send a human with a tiny welding rig out and tell them to hack. In situations where you'll never going to want to produce that precise artifact again (the exact way you reassembled the parts of this particular old bike frame won't be useful to someone wih a different one), analog production makes a lot of sense.

It's a setup vs execution time tradeoff - digital is heavy on setup time, analog heavy on execution. It's the same reason why software testing is done with automated scripts and manual exploration (see the work of John and James Bach, and Cem Kaner, and the context-driven school of testing, and their papers on exploratory testing). You pound on software manually because doing that is far more agile than writing scripts to do the tests you're thinking of (and 99% of those scripts would end up being useless and discarded after the first run, in any case). You write scripts because there's something you know you want to test the same way every time, many times. There's a sort of complex dance between the two - where best to use a person, where best to use code.

And call me sentimental, but there's just something viscerally satisfying about marking off and sawing wood, or shaving off plastic with a dremel, or feeding aluminum stock into a bandsaw and feeling the material respond to your movements. There's something in the way it feels to heft a good drill that can't be expressed in any coordinate system that I know of. We don't need to digitize all information. Not all information will be useful to digitize. And in any case, the end result (right now, at least) has got to be non-digital to actually physically impact the world.

What we can do is digitally share the results of these analog, human-driven adaptations. Pictures of "well, here's how I built my treehouse around this tree," an essay on "I took these old bike frames... and made a cart!" What we can also do is use digital fabrication to take the gruntwork away from analog fabrication so that people can spend their time in the creative bits ("how do I build a treehouse around this tree? Maybe I can cantilever this plank out from that branch...") and less time in the repetitive ones ("great, now I have to make a couple hundred treehouse-plank-holder-together pieces.")

What we can also do is push digital fabrication forward so that analog fab is forced to find a better and more optimal place to be (like how the advent of high-level languages freed low-level ones like assembly from having to handle All Computer Programming Problems and allowed practitioners of low-level languages to focus on interpreters, performance tweaks, and things that only low-level languages did well - admittedly, this isn't a great example).

This would probably make an interesting debate to blog out somewhere - digital vs analog fabrication, and where each method finds its strengths and weaknesses. Preferably when it's not 3:07AM and I'm half-zonked into lack of coherence from exhaustion.