It started two years ago, on a street in Dublin. They have a cool program where notable historical figures get a little bio on walls.
The key part was this:
She invented a unique uncrushable linen fabric by backing handkerchief linen with taffeta. Jacqueline Kennedy wore a pleated skirt in this fabric for her portrait by Aaron Shickler that hangs in the White House.
It must be this portrait, which doesn’t so much show a pleated skirt without wrinkles as suggest ethereal gauziness.
I’m working on a new outfit for my upcoming talk at ETLS, because what better reason for sartorial celebration than a book launch? (Pre-order Progressive Delivery!) I’m in a linen phase, and as I reviewed my interfacing options, I realized that I had a bunch of silk organza that I could combine with the linen to try this “uncrushable linen” theory. Here are the fabrics that will show (minus the organza):
The point
When I took that picture, I hadn’t sewn anything with linen. It was an interesting sign that I took a picture of on a day when I took hundreds of pictures. When I was in a circumstance where the knowledge would be useful, I remembered enough of the keywords to be able to retrieve the data.
Knowledge and learning are never wasted. They are delightful for their own sake, but also because sometimes things you learned for no reason can suddenly activate, like a sleeper agent that encounters the passphrase. The hard part is being able to retrieve that half-remembered concept you saw years ago. In the ideal form of SEO, and all types of information architecture, that’s what we’re facilitating – the ability to use a couple keywords and a dim memory of where you saw something to go back and retrieve it*.
I know we valorize focus and deep knowledge of specific things, and find tangents distracting, as a culture. But I want to say that I have learned as much about technical writing from knitting patterns as I did in any class. I want to say that I never explicitly use statistics, but I use them all the time. Nothing you learn is a waste.
* I think that’s why I have two search modes online. In one, I’m asking a straightforward question: “Juki serger 3-thread overlock threading diagram”. In the other, I’m trying to recapture enough of the context to find something that I am pretty sure exists, using whatever I can recall about it – location, company, season, keywords. “gmail:inbox delta las receipt”. Both of those searches are mostly deterministic – I expect that I won’t be bothered with Babylok serger instructions or Jet Blue flight itineraries.
The rise of LLMs that do semantic proximity is messing with that expectation. They know that Delta and Jet Blue both occur in the airlines semantic space, and they will try to be helpful by returning anything that matches that token, whether or not it’s what I asked for. This is not entirely bad – maybe I did need all my itineraries. Maybe there’s serger-threading advice that I’m missing out on. But a lot of it just feels like pollution of my fragile memories as I try to retrieve something.