For 2013 and some of 2012 and 2014, I mostly stopped knitting and spent the time teaching myself how to really make software. It wasn't exactly that intentional in the moment, but time is finite and both knitting and coding require similar kinds of attention and thought. In retrospect this is what happened.

I've started knitting again, not as much or in the same ways that I used to, but there's a bit more space in my brain for additional projects. I've thought often about writing about similarities between knitting and programming.

A sweater ontology

A photo posted by Sam Kleinman (@tychoish) on


The two activities have much in common. Code and knitting are both highly structured activities, with a reasonably finite syntax of possible component operations. Most programming languages have a few dozen keywords, fundamental types, and built in operations that give rise to all programs (more or less,) and knitting similarly provides a small number of operations (knit, purl, increase, decrease, slip) that you can combine and alter to produce pretty much any kind of fabric.

Knitting patterns are procedural (like programs, generally) and have a standard syntax that shares a number of common elements with contemporary programming languages, including loops, conditionals, map (functional) operations, as well as some non-trivial concurrency.

While each knitting stitch depends on the stitches that happen before and after it, they are all reasonably atomic within the knitted object. At the same time, many knitting projects have lots other objects that happen at the same time, along different dimensions: shaping, stitch patterns, short rows, sleeves/gloves as well as parallel objects like socks. Knitters routinely track multiple things happening at the same time, with some success: a task that many programmers struggle with.

The connection between knitting and programming isn't terribly surprising or novel: indeed there's a long history between the development of textile production and technology more broadly, but hand knitting is sort of a technological oddity, so the connections don't have the same history that you'd have in weaving (say.)

I've struggled to write this post for a while and have always had a hard time figuring out what the conclusion is or should be. I think I've felt like there should be some potential for synthesis, some way to take the things I've learned about software and use that to write programs to support knitting or design work, or vice versa. Increasingly, I think that this is the anti-conclusion.

Sure you can represent knitting patterns using formal languages (such as programming languages,) and it wouldn't be that hard to write software that converted a pattern, or validated the math, but the math is really the easy part of knitting.

The hard part about knitting things is not figuring out how many stitches you need to knit an object that's the right size, but rather figuring out what the right size is. It's easy to know how many stitches you need, and hard to decide how big or small a thing should be. It's also true that actually knitting things takes the largest amount of time; figuring out the math and making the design decisions are comparatively quick for most projects.

Computers wouldn't be able to tell how long we should knit the cuffs of our socks or the sleeves of our sweaters, no matter how much code we write.

In any case, I'll be over here, knitting away at my sweater.