Write Drunk Edit Sober

The famous quote write drunk, edit sober is often attributed to Ernest Hemmingway. When I first heard this, I thought of how I could incorporate this idea into my own creative process and came to the conclusion it is a terrible idea. As a creative professional, I need to come up with creative ideas every day. If I needed to rely on alcohol for my creative process, I would very quickly destroy my health. But something about this idea still rings true to me so I think it’s worth some time to analyze to see if there’s anything we can learn from it not just for writing, but for any creative activity.

Write Drunk

“I hate to advocate drugs, alcohol, violence, or insanity to anyone, but they’ve always worked for me.” – Hunter S. Thompson

The first part is write drunk. The first thing I think about when I hear this is one of my favorite authors, Hunter S. Thompson and the book Fear and Loathing in Las Vegas. Thompson’s inspiration for writing often comes from altered states of consiousness from the use of drugs and alcohol. Obviously you don’t need to go this far to be a creative person. But there is certainly something about the creative process that makes it feel like a different state of mind than normal waking consciousness. Creativity seems to flourish not by direct effort, but by the suppression of some more rational part of our personality that is responsible for inhibition.

The key insight is that the creative state of mind is sort of like being drunk. If we change this to write as if you were drunk the advice becomes much more practical. Someone who is drunk tends to be bursting with ideas. Most of the ideas are really bad, but there’s a lot to choose from. The drunk person has so many ideas precisely because he doesn’t care whether they’re bad or not. The alcohol suppresses the critical faculty responsible for their immediate evaluation.

We can mimick the creative part of the drunken state by slowing the feedback loop between idea generation and evaluation. An idea that seems bad may actually lead to some valuable path that we may have never discovered if we cut off the line of reasoning too early. For instance in chess, a common tactic is to make a move that intentionally loses a piece, sometimes even an important one like the queen, to gain a positional advantage that will win the game. If we consider the move and then immediately evaluate it, it might seem insane to intentionally lose our queen and the opportunity to win will be lost.

Edit Sober

Another great insight in this quote is that the creative process happens in two distinct stages. There is the drunk stage where you freely come up with ideas without judgement and then there is the sober stage where you pick one of the ideas and start to flesh it out. These two stages are completely different contexts, and switching between them incurs some overhead cost. Knowing when to switch is an important part of being creative.

With this process, the skill of creativity is to recognize a good idea through a process of selection. You become sort of like a music critic rummaging through recently released albums trying to find something to recommend to your readers. Sometimes a great piece of music won’t just jump out at you. Some of my favorite albums required multiple listenings for me to appreciate them. Many good ideas will challenge you to find their value. These tend to be the best ideas though, because if it were obvious, then everybody would be doing it already.

This selection process must be done sober. One of the problems with actually getting drunk is that drunk people make really bad decisions when it comes to selecting something to act on. I think we’ve all had that experience.

Applications to Programming

Since my craft is computer programming, I’ve thought about this quote in the context of what I do. Write drunk edit sober works for writing code too. When you first start on a project, none of the rules for best practices are practical. They just get in the way and slow you down. I make big ugly monoliths, write giant functions, hardcode everything, copy and paste big swaths of code around, and all the other stuff they teach you not to do the first day on the job. I can write code really quickly and efficiently this way because I’m self-taught and this is exactly what I did for the first few years and nobody told me any differently. I made some really beautiful disasters like this.

These days, nobody ever sees my code in this state because after I’ve gotten something basically working, only then do I clean it up and make into something pretty. After it works and all the details are in place, cleaning things up becomes really easy. The abstractions are neat and pretty because they were made at the last minute by necessity, not up front because of a guess. None of the cruft survives the editing process. Anything I show to anyone has probably been rewritten three or four times just by drunkenly iterating through bad ideas and then polishing up whatever is left.

I really wish language designers would take my workflow into account by providing me with tools to support both stages of work. I think we can split up programming languages (and frameworks) into those that are drunk and those that are sober.

Drunk Languages

  • JavaScript
  • Python
  • C
  • Scala

Sober Languages

  • Rust
  • Java
  • Go

The problem with this dichotomy is that it is very hard to write sober code in a drunk language and it’s very hard to write drunk code in a sober language. This is my main problem with Rust how it is right now. It’s extremely hard to drunkenly iterate with Rust code because it forces you to deal with a bunch of details you aren’t prepared to think about. In Rust, your code won’t even compile unless it’s guaranteed to be thread safe, has no memory leaks, and a lot of other things. Once you actually get your code to compile, it tends to be extremely reliable and safe. But your abstractions are going to be weird because it’s so painful to try new things without slogging through a bunch of details first.

A language like JavaScript has the opposite problem. Writing drunk code is really easy because there aren’t many rules. But cleaning it up after is very hard because your code doesn’t have a lot of enforcable structure to it. Anything can be anything at any time, which is very liberating in the first stage because it gives you a lot of flexibility, but becomes infuriating when you realize it’s nearly impossible to finish a big project cleanly. I’ve heard a lot of people complain about this when their Node projects ultimately become unmanagable when the structure becomes difficult to reason about.

I really wish someone would make a language that did both of these things well.

So I think the things we can learn from this quote are 1) don’t judge your ideas too early and 2) design your APIs for both stages of the development process.

Happy coding and please drink responsibly.

Leave a Reply

Your email address will not be published. Required fields are marked *