2025
🔗 Manuel Moreale: On em dashes
What if they tweak the instructions next week and tell it to use more full stops or commas? What are we gonna do then? Stop using those as well? Hell no. I’ll keep writing however I want, and if someone decides to stop reading what I write because they suspect it’s AI-generated because I use too many em dashes, or parentheses, or any other punctuation or word or whatever, well, good riddance.
My feelings exactly. To stop using em dashes because of AI chatbots seems ludicrous to me.
Aren’t the “remember me” checkboxes on login forms a bit unnecessary? I’m almost always using my own computers so of course I want the website to remember me. Granted, this is not the case when I’m using a shared machine, but maybe a “this is not my computer” checkbox would be more useful there.
I kind of wish I cared about the Apple event. I mean, I’m glad others do, and there was one time in the past where I did, at least vicariously (I don’t have an iPhone). But recently? 🤷 Might be that it’s a maturing product category, and that any advancements are incremental at best. Or maybe it’s just age, and being around long enough to have seen it all before. In either case, it’s getting harder to be excited about something that you yourself don’t use.
Apparently it takes Spotify to raise their price by $2.00 for me to cancel my subscription. I guess it acts like a good reminder of how much I was paying a month for a service I barely used. YouTube Music will be my secondary music service now (Alto is, and always will be, primary).
Ending the day on a win. Managed to diagnose and resolve a slow PostgreSQL query by creating an index, and also managed to configure Istio to allow communication between two Kubernetes cluster. These are super esoteric, and I’m not expecting anyone to care. Still, it’s good to leave work on a high.
I can’t for the life of me get this Docker image updated. I’ve tried deploying it to Kubernetes multiple times and it’s just refusing to launch any version newer than 4 commits ago. I may have to deploy via CI/CD, which will work, but is much slower than from the command line. ⏳
Really enjoyed this discussion on the latest Shoptalk show discussing why one would want a website. I disagree with the idea that restauranteurs can get away with just an Instagram. Restaurant websites generally suck, but I'm not sure what I'll do if they aren't around, and I can't see their menu.
Oh, and P.S. I don't know how to use Instagram.
Finally indeed.
I’ve been looking forward to this as a general train enthusiast. Then I realised it can improve my commute home. Now I’m really looking forward to this.
Speaking of Dequoter, in “celebration” of the upcoming release of MacOS 26, I styled the command palette a little, adding a transmissive blur to give it a glass effect:
It’s a shame I can’t style the options of the select element. It would be nice to increase their margins a bit.
Ah, MacOS’s locked-down nature strikes again! Was testing the CI/CD build for Dequoter and after downloading the artefact and attempting to open it, I got this warning message:
Turn’s out it was being quarantined by MacOS, and these instructions resolved the issue.
The binaries not notarised so I wasn’t expecting it to work out of the box. I was hoping that it would do that thing where the app will be listed in settings and I can allow it to launch from there, but I guess there’s something about where this file came from that was too much from MacOS. Ah well, I can live with this for the short term.
Father’s Day over here. Saw a family in the cafe with the father and son wearing matching Bluey T-shirt’s, with the message on the son’s T-shirt saying “My Dad’s awesome” and the Dad’s saying “Dad goals.” Like that idea.
🔗 HN Comment by IanCal in a discussion about RDF (emphasis added):
Someone will suggest modelling to solve this but here lies the biggest problem: The correct modelling depends on the questions you want to answer. Our modelling had good tradeoffs for mapping academic citation tracking. It had bad modelling for legal ownership. There isn’t one modelling that solves both well.
That may be why I was turned off by RDF all those years ago. One seeks to model a domain, but domains can be extremely complicated, and even if you cover everything, it’s still only one domain. But the biggest crime is assuming that the model is necessary for all use cases. And it just isn’t. Does every website that tracks books need to know the full legal name, publisher name, the legal entity of the company that supplied the typesetting? Simplify, man.
Via: Simon Willison
Devlog: Dequoter — Something Different Today
A new project called Dequoter was started to unquote a JSON string and filter it, utilizing Go for backend functionality and HTML for the frontend.
The 21st century - current progress:
████▒▒▒▒▒▒▒▒▒▒▒▒ 25.75%
It’s funny how the prevalence of electric cars seemed to snuck up around me. Four or five years ago, the number of times I see someone drive an electric car on any given day may have occupied one hand. Today, I’ve been outside for about an hour and I’ve already seen three drive by.
🔗 Dave Winer: We Make Shitty Software
We know our software sucks. But it’s shipping! Next time we’ll do better, but even then it will be shitty. The only software that’s perfect is one you’re dreaming about. Real software crashes, loses data, is hard to learn and hard to use. But it’s a process. We’ll make it less shitty. Just watch!
It’s true. Speaking for myself, I too make shitty software. Probably have my entire career. It’s only today that I’ve internalised it. And it’s a hard thing to admit. How hard? Well, try four attempts at posting this declaration publicly.
Via: Coding Horror
For anyone else that has unwanted YouTube channel subscriptions that they cannot unsubscribe from on the channel page because the subscribe button requires payment (that seems like a bug, Google), you can remove them by visiting the all subscriptions page (source).
I saw someone online mention Zo Computer so I though I’d give it a try. Asked it to produce a Go function based on one of my blog posts, since I needed the same thing in a different project. I saw it open the blog post and generate the function in about 5 seconds.
Granted that this is hardly groundbreaking. It’s using GPT 4.1 mini, so it’s likely I could’ve done the same thing straight from ChatGPT. But I think it’s a good first step in seeing what this service is capable of.
I also wonder if the model is actually consuming the post. I have nothing to support this other than doubt that the post would be in GPT 4.1’s training data. The URL can’t be more than a month old.
In 2021, I organised for a static IP address that can be whitelisted in the VPN of my previous job so I can work from home. Today, that same static IP address seems to be constantly added to the blocklist of the VPN of my current job. Yeah, VPNs are up there with DNS on the corporate pain-o-meter.