For reasons that don’t matter I’ve been watching a lot of Marvel movies I’ve already seen dubbed over in a language I don’t speak. It’s ok. I’ve found myself laughing at jokes even if I don’t understand the delivery, because I remember the joke from when I heard it the first time.

The whole experience is not unlike watching AVENGERS: ENDGAME in English.

A name is a perfect code, meaning exactly the thing it means and nothing else. In computing, often the name and the thing named are indistinguishable or the name of a thing inheres in the thing itself. We can look things up with identifiers, use them as passwords or as proof. If two things have the same name, then they are the same thing.

What are the gates and borders in your system that require names? What is the universe of naming? What attributes are part of the identities of your system, and what is incidental to identity?

There are a lot of animating myths about computers - computers are industrial, maybe they’re futuristic, maybe computers are tools of war or inhumane bureaucracy, computers are consumer products, maybe cyber is totalitarian or maybe cyber is punk or maybe it’s both. We use a lot of different stories when we make decisions about how much computer should go where, and why.

One story that has given us a lot, and seems to get a lot less credit than it should nowadays from you kids with your blockchains and your user experiences and your venture capital is that computers are psychedelic. The story goes like this:

A person with a computer, just like a person with language or writing, has a magical, prosthetic mind. An individual can use a computer to understand and know and explore things in ways you couldn’t understand and know and explore without one. That is what computers are for.

Here are a few useful technologies that sprang from that fountain, or at least drank deeply from it:

• Smalltalk

• The Apple II

• BASIC

• Hypercard

• Perl1

• Zork

• The Mother of All Demos

• ELIZA

• EMACS

• The Oregon Trail

• Web comics

• Wikis

• PROLOG

• The Hitchhiker’s Guide to the Galaxy

• Wendy Carlos

• Netscape

• Mondo 2000

• Demoscene

• Hacking

• The Mandelbrot Set

• Where in the World is Carmen Sandiego?

• Centipede

• The DNA Lounge

• Project Xanadu

• IRC

• Spacewar!

• Sketchpad

• String Art

• The fiction of Stanislaw Lem

• Bob Moog

• The Macintosh 128K

• TRON

• NeXTSTEP

• Objective C

• Yar’s Revenge

• The Spice, Melange

• Ultima

• The Later work of Jim Henson

• SHRDLU

• Magic Eye

• HAL 2000

• Della Derbyshire

• Who Framed Roger Rabbit?

• DEVO

• Jerry Cornelius

• MUDs

• Earthbound

• X-Screensaver

• and of course, the Tarot, one of the simplest psychedelic computers we can construct.

Notably not in this list are missile guidance systems, actuarial software, wherever they keep your permanent academic record, high-frequency trading systems, the iPhone, Uber, or Bitcoin.

In computing, the suit of wands represents the way computers transform information by encoding and decoding it. In the Tarot, the tenth card of a suit is the essence of that suit fallen into exhaustion, overwhelmed and used up: for computing, the 10 of wands represents undefined: non-information that lives where we think information should live.

Undefined is not the same as zero or NULL or a known unknown - and undefined value is an error to access. It is a book that is not in the library. It is a mistake, looking for information in a place that isn’t any place at all. If the answer is undefined, then the question is unintelligible.

What are you assuming about the structure of the problem that might not be correct? Where is the nonsense or misapprehension in your current model? What questions are you asking that don’t really have answers?

The 13th card of a Tarot suit shows us a teacher, and in computing tarot, that teacher is a person who teaches you about the reflexive, universal nature of computing. For the suit of swords, that teacher is The Mathematician.

Computers are a blunt imitation of mathematics. A lot of the development of Mathematics has been motivated by engineering and mechanism, but for digital computers, this is the reverse: the mathematics preceded the mechanism. Behind and beyond and before computers, motivating them, there is a truth. We learned about hypothetical computers as objects of pure mathematics, and then tried to build them in reality.

There is a human element that is easy to miss in the paragraph above - the “We” in “We learned” and “we tried”. The 13th card is the queen of that suit, a person, with desires and motives and limits. Why does the mathematician care? Why does she work? To what end?

To see a spark of magic inside of a computer is to be a programmer. Throughout the brief history of computing, the popular imagination of what and who a programmer is keeps changing, but the core truth the programmer - to understand computing as a verb, something made, to understand that making as a journey, to understand oneself as lost and found in how machines behave and surprise us - that hasn’t changed very much.

Why do you feel so strongly about the operations of a machine? What journey brought you to computing? How will it unfold?

In traditional Tarot, the last of the trumps is The World: the blessed completion, all things in harmony, the fulfillment of divine plan. For computing systems, the blessed completion we hope for is understanding.

Perfect apprehension of the system and its work is impossible, but better understanding is the focus of our effort. We can learn about systems in general, we can study this system in particular, and most importantly we can build systems in ways we think will help us understand them. Our end goal is the intelligible, the legible, the simple.

What are the tools for understanding you can bring to bear on your system. How can it be designed to be understood? What don’t you understand about now? What are the limits of your understanding?

The Tarot suit of pentacles, for computing, is the physical dimension of computing: computing in space, in time, with weight and electricity and heat and manufacturing and supply chain. The Ace of Pentacles is manifestation - the mechanical reality of computing.

Computers are mathematical but not mathematics. When we theorize about computers, we theorize about their limits: how much time a computation will take, how much memory will be needed. Computers in reality use electricity and produce heat. Setting up the air conditioning for data centers is an important and complex art.

The cards of the suit of disks are all aspects of the physical, finite nature of computers - where we’ve already drawn one of them (and disks seem overrepresented in the history of this deal) they are linked:

1. The Ace, the manifest property of computing itself.

2. Time complexity

3. Memory

4. Glitch

5. Overflow

6. Process

7. Reordering

8. Heat

9. Administration

10. Crash

The court of disks are working engineers, engaged with the limits and realities of computing.

• Princess of Disks: The Network Engineer

• Prince of Disks: The Build Engineer

• Queen of Disks: The Security Engineer

• Knight of Disks: The Computer Operator

C contains a very clear, fairly simple model of a computer. In 1980, this model was fairly accurate - today it is the story we tell one another to make sure we are understood.

C’s design centered simplicity, here understood as simplicity and intelligibility of implementation. Because C was portable, intelligible, and efficient, it became ubiquitous. Because C is ubiquitous, it is the lingua franca of programming, and the foundation of many, many other technologies.

C programmers have a reputation for pragmatism, for valuing efficiency, for making an effort to completely understand the implications a decision. A C programmer values simplicity and mastery. A good C programmer is a great system designer, and has a depth of understanding of their system’s environment, assumptions and dependencies.

An obsession with mastery can breed contempt for non-masters, and a belief in one’s on mastery can lead to arrogance at best and cruelty at worst. “Simple” decisions about C memory management, pointers, arrays and string handling, and the belief that master programmers can work safely amid those decisions, have cost humanity billions of dollars and probably more than a few human lives. C is dangerous, and in reality there are none of us can actually master the complexity of a real, life-sized computing system.

In the standard Tarot, the 20th trump card is The Last Judgement. For computing, the 20th trump is Commentary, the human language documentation and notes that accompany a computer program or system.

To understand computing systems, we rely on messages and guides from folks who have come before us. Architects draw diagrams we consult. Authors of programs write their reasoning next to the results of their decisions. What is illegible from the system itself is instead written in its margins.

Of course, commentary is unreliable in ways that the source code of the system itself is reliable: while comments tell us what the author intended, the source code tells us what is actually happening. Comments are wishes that we hope align with reality. As systems change, comments that once helped can lead us astray, or just distract us from the real things in front of us.

How do the stakeholders in your system communicate? What are the artifacts of that communication? How long do they live? Long enough? Too long? How does the society of your system align with the system’s mechanical operation?

Partitions may not be permanent, and may not be obvious. A common side effect of partitions is a split brain, where both parts of the network believe they are the whole. When the partition is repaired, which side of the repair represents the “real” history of the network? How can we reconcile the two experiences?

Where is the heart of your system? If your system is well designed, that heart may be in many places, or everywhere. If that is the case, what happens when those hearts differ?

The Internet is less a technology than a set of agreements about technology: how systems will behave with one another, and respond to common messages and circumstances. The same is true of many programming languages - there are many C compilers, but they all agree to interpret the same language, C. C is an agreement, held for the convenience of compiler writers and users.

Standards are arbitrary - no divine voice says that IP addresses should have 32 bits instead of 64 bits, it’s just an agreement some folks made a while back. That said, once made, and built into systems, these agreements can become very, very difficult to change.

What are the basic expectations your users and peer systems have about your behavior? Are they codified? Can they be codified? Are they over-codified? What will you be able to change later, and what will you have to live with forever?

In the traditional Tarot, the 10th trump card is the Wheel of Fortune. For computers, it is the copy.

Information has a location, but those locations are ephemeral and changing. To use data from your disk, it must be copied into memory; when you retrieve a document on over the Internet, you reproduce the entire document on your own computer. Our intuitions about identity start to fail under these conditions - is there an original version of this or that information? How does it differ from the version I have in my hands? Does it matter?

Consider matters of identity in your design. What are the implicit differences between nodes, services, transactions? How can these differences be eliminated? How must different parts of your system remain in sync?

The princess of wands is the Smalltalk programmer. A Smalltalk program is composed of many small, opaque servers that receive and respond to messages.

Smalltalk is total, utopian, orthogonal, uniform, elegant, full of wonder. It assumes no difference between using a computer and programming a computer, it was designed to be taught to and learned by children. Smalltalk was a central part of the original vision of “Personal Computing”. All programming languages are a vision, but Smalltalk says “Smalltalk is a vision.”

A Smalltalk program disintegrates into tiny methods against stacks a million frames deep. The language is tiny but the Smalltalk system is sprawling and enormous and all comes from the same place. Smalltalk systems tend to the incomprehensible - you cannot use them a little bit, you have to use them in their totality.

As you simplify one part of your system, where does the complexity go? How will your system fit in with other systems, driven by other users and use cases you don’t yet understand? What are the unifying metaphors or abstractions? What is complete and what is partial?

Barb and Star reads like Twin Peaks, if Twin Peaks was earnest and hilarious and sweet instead of vaguely angry that you chose to spend your time watching TV. This movie is amazing, and you should watch it.

Barb and Star is movie a constantly threatening to destroy itself. Almost every decision in the movie is some refutation of the way this particular movie moment is supposed to to work. The movie is constantly playing chicken with its medium.

Inside and around the madness, though, are people we know and like - the pair of friends and the center of the movie aren’t real, exactly, but they’re real enough, in ways that make you care about them and understand them, and follow along laughing when the movie does a pirouette at the edge of a cliff or unwinds or otherwise refuses to behave like a story. The movie plays chicken with its medium, and the movie wins.

The suit of swords, in computing, represents the reflexive and universal nature of the computer. The fifth card in a Tarot suit is the essence of the suit places that suit as an intruder, challenging and disrupting the world. The five of swords represents recursion - the ability of a computer program to invoke itself - that magic trick and cornerstone.

def find(needle, haystack)
  if haystack.empty?
    false
  elsif haystack[0] == needle
    true
  else
    find(needle, haystack[1..])
  end
end

For recursion to work, it must have a base case. Each recurrence must be for a smaller problem than one before, converging on an answer. If the problem doesn’t get smaller, then you have induction, diverging out into infinity: that works for mathematics, but not for our finite computers.

Consider the definition of the issue you’re encountering. What are the trivial cases or zeroes? What are the next steps up the mountain? How will you know when you’re finished?

ENIAC, “the first programmable, electronic, general-purpose digital computer” was built to calculate ballistics tables. The Defense Advanced Research Project was the funding behind the first great age in the development of computers. The Internet was designed as a way for computing infrastructure to withstand nuclear attack. The great engine behind the early development of computation was war - cataclysmic, all-burning, 20th century great-powers war: the coming war that no one would win.

We continue to try to hammer these swords into plowshares and cellphones and video games, but the legacy of war is part of every bit of computation. To build computation for peace we have to encounter and work around and subvert and rewrite the logic and history of war.

Confront the dark parts of your history. Build good things, and admit and work with the possibility of their bad beginnings. Put down your shoulder and push away from the trajectory you have been given, towards something new, and something that is yours.

In computing, the suit of wands represents encoding - the way we pull information out of wires and make information out of alphabets and the ways we get information from one place to another. The fifth card in a tarot suit is an overreaction to the trauma of the nature of the suit when it encounters the world. The fifth card of the suit of encoding is noise. Noise is the difference between what the encoder intended and what the decoder receives.

Noise is something to be worked around. Noise is unpredictable in the particular but is predictable in aggregate, so we can communicate in the presence of noise with redundancy - we repeat the same thing three times, and choose the two that look the most alike. We measure again and again and take the average.

Consider the well understood frictions and obstacles in your path - the little work-arounds and blockages in the flow of your effort. Consider how surprises even out at scale, and whether or not the sum of your mistakes goes to zero in the long run.

Computers are “universal machines”, in that one computer can simulate any other. The suit of swords represents the reflexive quality of computing.

The fourth card in a tarot suit represents the shock of that suit’s core quality, dropped out of the ideal and into the chaos of the real world. The motivating crisis of the universal, reflexive quality of computing can be seen in the halting problem: because computers can simulate one another, we can prove that certain kinds of very useful programs are impossible to write.

When Alan Turing wrote the paper that described the halting problem, he also produced what we think of as the first description of a modern computer. In the service of proving that some programs were impossible, he had to invent programming.

The details of the halting problem are complex for this short message, but not too complex for non-specialists to understand. Turing’s paper has a scary technical title, but like a lot of foundational science papers, it is relatively easy for non-specialists to read because there weren’t any specialists around when it was written. It’s worth a look.

How are constraints structuring your thinking? How can what is impossible inform what is possible?