The difference between "low prices" as a race to the bottom or as a success story (like Amazon, Costco, IKEA, Vanguard) is in leveraging intentional weaknesses.

I’m trying something a bit different today – fiction. I had an idea for a short story the other evening, and I fleshed it out into a proper piece. I want to get better at writing fiction, and the only way to do that is with practice. I hope you like what I’ve written! When the fire starts, I am already running for the exit. When the fire starts, the world is thrown into sharp relief. I have worked in this theatre since it opened its doors. When the fire starts, my work begins – and in a way, it also ends. When the fire starts, they run beneath me. When the fire starts, they leave their bags behind. Their coats. Their tickets. They hear me, though I have no voice. When the fire starts, I know I will never leave. When the fire starts, I will keep running. I will always be running for the exit, because somebody must. A “running man” exit sign. Photo by Mateusz Dach on Pexels, used under the Pexels license. Hopefully it’s clear that this isn’t a story about a person, but about the “running man” who appears on emergency exit signs around the world. It’s an icon that was first devised by Japanese graphic designer Yukio Ota in 1970 and adopted as an international symbol in 1985. I was sitting in the theatre on Friday evening, waiting for the second half to start, and my eye was drawn to the emergency exit signs. It struck me that there’s a certain sort of tragedy to the running man – although he guides people to the exit, in a real fire his sign will be burnt to a crisp. I wrote the first draft on the train home, and I finished it today. I found the “when the fire starts” line almost immediately, but the early drafts were more vague about the protagonist. I thought it would be fun to be quite mysterious, to make it a shocking realisation that they’re actually a pictogram. I realised it was too subtle – I don’t think you’d necessarily work out who I was talking about. I rewrote it so you get the “twist” much earlier, and I think the concept still works. Another change in the second draft was the line breaks. I use semantic linebreaks in my source code, but they get removed in the rendered site. A paragraph gets compressed into a single line. That’s fine for most prose, but I realised I was losing something in this short story. Leaning into the line breaks highlights the repetition and the structure of the words, so I put them back. It gives the story an almost poetic quality. I’ve always been able to find stories in the everyday and the mundane – a pencil is a rocket ship, a plate is a building, a sock is a cave. The only surprising thing about this idea is that it’s taken me this long to turn the running man into a character in one of my stories. I really enjoyed writing this, so maybe you’ll see more short stories in the future. I have a lot of ideas, but not much experience turning them into written prose. Watch this space! [If the formatting of this post looks odd in your feed reader, visit the original article]

As part of my work on #eng-strategy-book, I’ve been editing a bunch of stuff. This morning I wanted to work on two editing problems. First, I wanted to ensure I was referencing strategies evenly across chapters (and not relying too heavily on any given strategy). Second, I wanted to make sure I was making references to other chapters in a consistent, standardized way, Both of these are collecting Markdown links from files, grouping those links by either file or url, and then outputting the grouped content in a useful way. I decided to experiment with writing a one-shot prompt to write the script for me rather than writing it myself. The prompt and output (from ChatGPT 4.5) are available in this gist. That worked correctly! The output was a bit ugly, so I tweaked the output slightly by hand, and also adjusted the regular expression to capture less preceding content, which resulted in this script. Although I did it by hand, I’m sure it would have been faster to just ask ChatGPT to fix the script itself, but either way these are very minor tweaks. Now I can call the script in either standard of --grouped mode. Example of ./scripts/links.py "content/posts/strategy-book/*.md" output: Example of ./scripts/links.py "content/posts/strategy-book/*.md" --grouped output: Altogether, this is a super simple script that I could have written in thirty minutes or so, but this allowed me to write it in less than ten minutes, and get back to actually editing with the remaining twenty.

Building a mental model of computer architecture from first principles

One of life's greatest simple pleasures is creating something just for yourself.