Adrianna Pińska

Random blast from the past: in my sewing stuff I found a folder of handouts I made for a local Society for Creative Anachronism class about the St Birgitta's cap. It's just one double-sided page -- here are the grayscale PNGs, and you can get the PDF here: https://drive.google.com/file/d/1dhlGto7INi9ptzDEQYWLgR8rxFw0GIIo/view?usp=sharing

This is a nice beginner project. It's a small item, so it doesn't use a lot of fabric, and is feasible to hand-sew if you don't have a machine. It may seem like an obscure piece of historical clothing, but it's handy for getting your hair out of your face or off your neck if you've just impulsively cut it into a layered bob and it's hard to tie up. You know, hypothetically speaking. It also has historybounding possibilities -- you can use completely ahistorical modern fibres, colours, prints, etc.. Maybe do a silly one to use up some scraps in your stash.

#sewing #sca #historybounding #reenactment

I really really don't want to make complaining about dumb LLM stuff my entire online personality, but dumb LLM stuff keeps happening.

I regret to inform you that not only do some people not check whether the LLM-generated "summary" they get when they do a web search is correct; not only do they have zero comprehension of what they're even looking at and how it differs from the way that a web search worked ten years ago; they may *literally be unable to understand whether the summary has answered their question*.

This is a thing that I have actually seen with my eyeballs in the wild. A recent episode of a TV series I'm watching was dedicated to someone's memory. In a discussion forum, one poster wondered who that person was, and another poster helpfully provided the "answer" they got "from Google", which was clearly generated, and said something like: "Sometimes when a person involved with a TV production dies, the crew dedicates an episode to their memory. For example, a recent episode of $SERIES was dedicated to $PERSON." Zero information about the identity of the person. The first poster said something like "thanks, that's very interesting."

This is very cool and normal, and definitely isn't causing any problems.

A concrete call to action, so that this isn't just a doompost: if there are people in your life who are less technically-minded than you are, and who trust your technical expertise to some extent -- try to explain to them what is happening here, and how they can try to avoid it. Maybe construct a demo using a knowledge domain that they know very well. Maybe offer to configure their default search engine to strip out the bullshit. Obviously that doesn't solve the problem of the slop farms flooding the search results, but you could show them how to recognise a slop farm, and the "back in time" trick for finding pre-collapse content (not that that's foolproof, since some slop farms backdate their pages).

Please have these conversations even if your view of LLM features in search engines is more positive than mine (that's a low bar). I hope that we can at least agree that people should make informed decisions about these tools, and that is not something that we can take for granted.

I fixed another long-time annoyance today: the missing window icon of the Slack desktop app (also missing from my pager). This is a common problem with applications that come from outside the traditional Linux ecosystem and are a bit half-assedly integrated, especially with pre-XDG standard environments. Like Fluxbox, which is what I use.

The issue is that the app doesn't set a `_NET_WM_ICON` property on its window. In an environment that supports XDG, you can work around this by setting properties in the `.desktop` file -- so if you use one of those, don't do this; do that instead. But Fluxbox doesn't use those files. There are various scripts that can generate a Fluxbox menu from the XDG application tree, but it's not a native feature. So you need something older to apply the property to the window at runtime.

There's a utility called `xseticon` (this one https://github.com/xeyownt/xseticon, not the other one) that does this (converting an input PNG icon to the appropriate format). You can install it as a snap, but then you have to give it special permissions to access icon files outside your home dir, so I suggest building from source (it's straightforward).

`xseticon` needs the window's name or ID. There are multiple tools that can find this for you. I picked `xdotool search`, because it has a `--sync` flag that makes it block until it has output, so I didn't have to hack in my own solution to wait for the Slack window to be found immediately after it is launched. I had to combine a bunch of search parameters to find the correct Slack window and not any of the incorrect ones (searching for visible windows up to a specific depth narrows it down enough).

Then I wrote a wrapper script to launch Slack and fix the icon, and put it in my `~/bin`:
```
/usr/bin/slack "$@" &
WINDOW_ID=`xdotool search --sync --maxdepth 2 --onlyvisible --class slack`
xseticon -id $WINDOW_ID /usr/share/pixmaps/slack.png
```

Note that if you call the wrapper `slack`, as I did, it's EXTREMELY IMPORTANT that you call the real Slack executable with its full path and not an unqualified name, unless you want to fork bomb yourself and crash your computer. I didn't do it this time, because I did it the last time I made a wrapper like this.

#linux
1/2

Parts of Cape Town are experiencing severe flooding, and this is hitting poorer areas especially hard. Now is a good time to donate some money, if you have some to spare. Decision paralysis? My suggestions:

Gift of the Givers -- they're my go-to organisation for disaster response donations. You can make a general donation or specify where and how you want the money to be used.
https://giftofthegivers.org/donate-online/

PEP Stores Lay By Buddy -- not disaster-specific, but a lot of people are going to have to replace a lot of clothes and other essentials.
https://www.pepstores.com/pages/lay-by-buddy

Cape of Good Hope SPCA -- they don't just deal with strays; they also provide emergency and veterinary assistance to people who can't afford to pay for it. They're currently also rescuing farm animals (whose loss would have not just a psychological but also a financial impact on their people).
https://capespca.co.za/flood-crisis-response-be-their-lifeline/

The LLM word vomit has spread to online cooking instructions, which is exactly what you want to see when researching how to prepare a new kind of seafood without getting food poisoning.

I recently listened to the Well There's Your Problem episode about the Therac-25. It's from a few years ago, so it predates the current moment of mass psychosis in the software world.

I immediately wondered what the new "AI"-enabled Therac-25 was going to be, and then I remembered that we already have at least one, and probably more that we don't even know about: https://www.reuters.com/investigations/ai-enters-operating-room-reports-arise-botched-surgeries-misidentified-body-2026-02-09/ (CW for medical errors and body horror).

Re-reading this article after listening to the podcast is chilling because of the parallels: the lack of effective regulation, the repeated denials from the company that anything is wrong, the lawsuits... (two of the Therac-25 incidents [that we know about] led to lawsuits that were settled out of court). It may be years before we know the full extent of the harm caused by tools like this.

Yes, there are valid use cases for machine learning in medicine, some of which are safe. Guiding a surgeon inside a patient's skull clearly isn't one of them.

What can anyone do about this as a patient? Ask hospital staff whether and how machine learning models are used in the software they use? How can they possibly answer that? Do your own research? This is all proprietary tech, so good luck with that. That's if you're even in a position to do anything; you could be going into emergency surgery, unconscious. Do we need "NO SLOP" medical bracelets? See the first point. Horrifying.

I don't have a solution, except to hope that the bubble bursts and all these clowns go under before their products make it to where I live.