Fifteen years of conversationsAug 6, 2017
In high school, I began saving my chat logs on a hunch that they’d be useful to look back on later. Fifteen years later, I’ve finally got around to parsing and importing them into my Memex.
One of the most amazing things about being a teenager is how much time we had to just sit around and do nothing. A lot of the conversations reflect that (I’m Kermit):
It’s all there: the awkward crushes, the anxiety about not being able to cut it in university, the major shifts in beliefs. There are the very stark reminders of the things I got wrong, like the year I became a libertarian and harassed my friends into doing that online quiz that would show that they too should become libertarians. I’ve spent a lot of time cringing this week.
Some of the conversations were about the logging itself. I made a joke with a friend about getting together in ten years and reading the transcripts (we never did 🙁). Other friends weren’t as excited as me about the idea of logs: “so do u like save every conversation? FREAK!” And some of the transcripts reflect a premonition of the annoying technical problems I’d have to solve to parse them properly:
In my personal Memex, I now have an almost a complete record of everything I’ve typed for the last 15 years. The majority isn’t worth ever reading again but there are many moments that make it worth it.
But the real goal of this parsing/importing project is to build two chat bots, one trained on chat logs of high school Andrew Louis, the other on contemporary Andrew Louis, and get them to talk to each other.
An attempt to import Goodreads data
In general, the internet is a dumpster fire of bad behaviour. Goodreads is a pleasant counterexample, a place where people have civil discussions about books. I’ve used it for years to track my reading and get recommendations and I know it’s popular with a lot of the people who I know will want a personal Memex.
In theory, building an importer to sync in book activity (read, currently reading, to read, etc) should have been a pretty straightforward exercise. Unfortunately, the API hasn’t had much love recently and I ran into all kinds of problems. Most seriously, I couldn’t even get consistent results when loading the book shelf endpoint — I posted about it on their forums but haven’t heard a peep yet.
The importer is stalled for now. West Coast friends: go work for them and fix these problems!
Goodbye, automatic screenshots
Out of all the categories of things I’ve started logging, I’ve only discontinued two. The first is calories. it was too much work to guess calories for each meal and I wasn’t even convinced I was always within the right order of magnitude. Even in perfect lab conditions, there are lots of big questions about their usefulness as a metric.
This week, I discontinued the second. For years, I had a script that would take a screenshot of my laptop every hour. They were sometimes nice to look back through but almost everything visible in the screenshots were tracked in other ways already (browser history, code, chat windows) and 5MB+ screenshot images take up a lot of storage space. Another problem: a huge percentage of these screenshots were just screensaver/lockscreen:
I’m going to be more intentional about taking screenshots of interesting moments or clips of things on my screen and I’ve set up an importer to automatically watch for new screenshot image files to import automatically.
I’ve always loved Joy Division’s Unknown Pleasures album cover (TIL: from a data visualization of pulsar) and I decided to try and build something roughly inspired by it. Using threejs, I made a mesh wave generator and got the camera fly over them.
It was fun to learn threejs and I hope to use it in the future to visualize some aspect of my Memex data.