Scrawlspace

This is scrawlspace. I scrawl in this space. Do not expect coherence or permanence…

…actually, looking back at the history for this part of the capsule, you can mostly expect permanence by this point, at least going forward. I used to rearrange things more often than I do now.

Normally I’d put a link to ../2026/ here, but 2026 hasn’t happened yet. You’ll have to wait AT LEAST for 2026 to happen. You might even need to wait for me to actually write something IN 2026 IN Scrawlspace.

2025-12-04: Finally, some good news

The merely good news:

Explaining the even-better news:

The Liquid Glass guy jumped ship and I wish him all the best at his new company.

If you’re a Halfway to Mars completionist you’ll know that I have had more exposure to more “dye” puns than I could ever stomach normally (and I mostly had the guy spiffing up my glider), so you’re not getting any here.

2025-11-07: Marathon wait, iOS edition

I read about what’s new in iOS 26.1:

This seemed interesting, and worth writing down because I’d never remember it otherwise:

There’s a hidden Apple Music gesture in iOS 26.1. You can swipe on the music player in the Apple Music app to switch songs. Just slide a finger over where the song title is listed, and it'll go to the next song or back to the previous song.

I’m still waiting another month+ for 26.2 (I usually only get a chance to upgrade my main machine over Christmas break), but this’ll be a nice reason to update the phone OS as well (iPadOS gets proper multitasking and macOS gets 24-bit color in Terminal.app).

2025-11-02: unc’s first Halloween

I thought about life, the universe, and everything for a bit and came to the conclusion that, unlike many previous years, I didn’t have any good reason to not hand out Halloween candy this year.

Also:

• I only want to hand out the kind of candy I would have wanted to get.
• There aren’t THAT many kids around me, right?
• Candy these days is made of more filler than there used to be unless you go for some sort of premium brand.

The question then becomes “do I want to accidentally have leftover candy, or do I want to risk running out prematurely?”.

I decided to err on the “no leftovers” side of things and bought a single 30-pack box of Hershey’s brand things. Plain chocolate, chocolate with almonds, Reese’s peanut-butter cups, and Kit Kat.

Of course, I wanted to taste everything beforehand. Quality control ain’t easy.

The peanut-butter cups and Kit Kat were great.

As for the chocolate bars…short story long†, while I have many happy memories of eating Hershey’s Kisses back when I was a kid, I’ve totally deacclimated to the butyric acid present in Hershey’s chocolate after buying and eating Belgian-made Pound Plus (500g) bars at Trader Joe’s for a few decades now.

I tried to keep the Hershey’s chocolate bars on top to make it more likely that I’d like to eat any leftovers, but kids would dig for the good stuff (my only instructions were “take one”).

I ran out mid-way through the night, but on a group-of-kids boundary, so it wasn’t the literal worst. The porchlight went off immediately after.

Next year I’m getting TWO 30-count boxes, and neither will have Hershey’s chocolate in it. I’m hoping to get some sort of individually-wrapped premium brand so I don’t have to worry about leftovers quite so much, but I didn’t notice anything like that a couple months ago. I think I’ll have to keep an eye out all summer long.

Totally related:

Kinda related:

⁂

† Short story short: bleaaghhhhhh…do my fellow Americans REALLY?

2025-11-02: In case you’re wondering, people ship software fast using AI, and they’re not stupid about it

People talk a lot about poorly-written code that LLMs spit out.

Of course, one could just…use tools like that with the normal amount of oversight that one would use for software engineering:

Some startups are already near 100% AI-generated. I know, because many build in the open and you can see their code. Whether that works long-term remains to be seen. I still treat every line as my responsibility, judged as if I wrote it myself. AI doesn’t change that.

>

There are no weird files that shouldn’t belong there, no duplicate implementations, and no emojis all over the place. The comments still follow the style I want and, crucially, often aren’t there. I pay close attention to the fundamentals of system architecture, code layout, and database interaction. I’m incredibly opinionated. As a result, there are certain things I don’t let the AI do. I know it won’t reach the point where I could sign off on a commit. That’s why it’s not 100%.

2025-10-27: I had no idea that this would be the new multiline twtxt

You’ve heard of twtxt, right?

You throw up a file, likely called twtxt.txt, on your web server and run this program to edit it and gather others’ single-line thoughts. There’s at least one Geminaut out there who’s doubled down on indie and has a twtxt feed sitting around on a capsule somewhere.

Of course, lots of people have multiline thoughts.

So someone made this:

So instead of uploading a file full of one-line posts, you upload an org-mode (it’s an emacs thing) file full of potentially many-lined posts, along with all kinds of metadata about the post.

I’m not an emacs guy but I thought it was neat to see that I’m not the only single-file (micro)blogger out there.

2025-10-20: Obsidian, now significantly freer beer

I was reading the Omarchy docs, as one does, and was shocked by this:

Obsidian is free for all purposes, including personal, commercial, and non-profit use.

after checking the Obsidian pricing page—

I was pleasantly surprised to see that they’re not so insistent on their pound of flesh anymore if you’re using Obsidian to make money.

You should still send them money if at all possible if you use their stuff, because they’ve got a great app and they’re _very_ committed to open formats (an Obsidian vault is just a folder full of Markdown files and a hidden folder full of JSON configuration). But they don’t insist on $8/month or whatever it was for a commercial license anymore.

2025-10-18 (started before I was totally done with the previous post): After an extended light Saturday-afternoon reading session

I changed my text-editing settings for my capsule to use a proportional font.

If you use a fancypants text editor, you might be able to have per-workspace or per-document font settings so you can still use monospace fonts pretty much everywhere else.

Sure, I’ll need to switch back at least briefly if I want to add in some nontrivially indented code in a triple-backtick block, but not using a monospaced font is downright refreshing even though my daily-driver monospaced fonts are very good.

Just sayin’.

2025-10-18: 100% − ε

I finally read that book I shilled:

I think I’ve had roughly four different ways of kicking off a proper review of the book (and not just a scattered collection of thoughts/notes like Wrangel’s _Always with Honor_).

I don’t really have extended thoughts instantly about it. I have one unique relevant anecdote, but I’m not sure I want to spin up an entire page in /words/ for it. I could puff up the page by mentioning an unrelated opinion, but I’d need to do a fair bit of work to flesh it out into a thought that makes sense to people who aren’t in _one_ particular inner circle of my friends.

Probably the hardest problem _I_ need to deal with that this book raises is “how much do I want to devote to this sort of problem”. I have loads of opinions on domestic policy, and as much as I think normal retail politics is focused on stopping normal people from doing bad things, I have the sinking feeling that this is essentially a distraction from much, much MUCH more important things that have a near-certain chance of rendering 2020s-era human politics utterly irrelevant — think of the effects the Maxim had on African inter-tribal warfare. Except the Africans as a whole ended up OK, generally speaking, compared to what almost certainly awaits us if a superintelligence gets made.

Not unrelated:

2025-10-13: It’s not an earwax problem

JP talks about lossless audio:

## Is it worth it?

To find out if it’s worth it for your ears as you are right now in your usual listening place with your usual setup, you’ll want to ABX your music:

I tried this a gazillion years ago and found out that Vorbis gets transparent somewhere between -q 4 and -q 5 on Weezer’s “My Name Is Jonas”. The default at the time was -q 3. -q 10 is the maximum, and anything above -q 8 prompts people to ask you why you don’t just use FLAC. At any rate, I’d use Opus if I wanted lossy audio and didn’t care about widespread compatibility.

I still use ALAC (in Music.app) and FLAC (everywhere else) because I like the warm fuzzy feeling of never having to wonder if I’ll ever outgrow my audio setup and listening skill. I think the big winner of this preference of mine has been Apple, because having all my music in lossless has caused me to bump up to the next storage tier for their computers and handheld gizmos more than once at $100 apiece.

And you know I always appreciate a good artist recommendation.

Have you heard (of) Tetsu Inoue’s “Ambiant Otaku”?

I can’t CONFIDENTLY recommend it based on “I like the Cyberpunk 2077 OST” (as well as descriptions of other artists I’ve never heard of) because I’ve never played the game myself, but the YouTube link is RIGHT THERE, so you can easily find out if it’s your cup of tea or not.

2025-10-04: They took the darndest things from you

Back in the oldest of old days, you could just wait for a computer to finish doing whatever it was doing, and then flip the power-off switch.

Eventually computers got complicated enough that you needed to select some kind of explicit “shut down” item in the operating system so the OS could clean things up properly and then shut itself off, or at least tell you it’s safe to shut the computer off.

Windows eventually gained enough smarts to start a polite shutdown sequence if someone presses the power button. (Holding the power button would still cut power, with ill-defined negative consequences down the line.)

However, Windows regressed. In order to speed up the boot-up sequence for normal people most of the time, a modern Windows machine will usually log in and start up programs while the machine is at the lock screen, waiting for the user to type in his Windows Hello PIN or full-fledged password or whatever.

…somehow, this means that when I press the power button on a Windows machine where nobody’s logged in, the machine doesn’t shut off anymore. Normally this wouldn’t be a problem, but for a while like a month or two ago I wanted to start up a Windows machine and shut it down using only the button.

2025-10-03: A very limited defense of RGB lighting on input accessories

You may have read this already:

The relevant bit:

To my own shock I'm using one of those really nice retro looking little 65% mechanical keyboards like all the cool kids use. [Many are] festooned with RGB backlights (which are an unconditional aesthetic travesty, if you like them your taste is bad and you should feel bad).

I have a Razer mouse (long story). Out of the box, the snake-king logo on the back cycles through all the colors of the rainbow.

This is, to put it politely, not my jam.

Luckily, the Windows controller software for it will let you turn it to one color, and will let you turn the brightness down quite a bit.

A Razer logo that’s dark green and is so dim that it’s only visible when all the lights are off and my eyes have adjusted to the darkness?

I never would have thought that this would be in the vicinity of my jam, but it is.

As for keyboards…

I chose a keyboard that doesn’t have RGB lighting (it’s bloat, and I was planning on plugging it into my iPad on occasion which has better uses for its battery than doing ambient lighting). That said, even though I’m a blank-keycaps kind of guy now, I’ve noticed that a teeny little bit of ambient lighting helps me put my hands on the keyboard and not miss. You’d think that you wouldn’t need to look at a keyboard at all unless you’re looking at what’s on the caps, but I guess that isn’t quite true all the time.

Dim red LEDs could be handy in the dark. Dim orange LEDs could be both handy AND festive in a couple weeks. Dim red-and-white LEDs could be festive in a couple months, but the white ones are going to wreck your night vision more than I might want.

2025-09-30: I should be looking around instead of typing on my phone

Luckily, phones today have voice recognition that isn’t awful. Even better, Textastic well automatically curl apostrophes even though that kind of thing generally doesn’t happen with voice recognition.

In order to make this post a properly OFFLFIRSOCH-compliant post, I plan on pushing the change to my computer from outside my house and SSHing into my computer from my phone. But that comes in like an hour or two from now.

I think the weird thing about today’s hike wasn’t the weather, or anything else about the hike itself, but me. For some reason I’m getting more distracted than usual. Usually when I go on a hike like this, in a place I know well, where I’m not with anybody else, I listen to podcasts. However, today I noticed that my mind was wandering so much that I wasn’t really paying attention to the podcast I was listening to, so I decided to stop playing podcasts and actually listen to music.

I ended up settling on Dvořák’s From the New World. For a hike like this, the introductory — I think it’s a “movement” — was too stormy and thunderous. Sure, it says “adagio, allegro molto” but I wasn’t really feeling the adagio part of it. When I get back to civilization I’ll have to look up the rest of the Italian and find out if it actually makes sense or not.

The second track, which I assume is a second movement, actually matches the vibe that I’m seeing and feeling right now. I think I’ll listen to some more of that right now and then come back to composing this later.

Yeah, that was the right call. Now it’s moved on to number three and the first word of that is “scherzo”. It’s harshing my mellow, so I’m back to not listening to that.

Turning the music back on, I have to admit that I very much do like the vibe of this particular movement and it’s significantly more familiar to me than the second one. That said, it doesn’t really match the environment around me unless the wind kicks up to the point where it becomes mechanically challenging.

Kinetic. That’s the word I’m looking for to describe how what I’m listening to isn’t a great fit for everything around me. It’s great music of course, but not really a good fit for what I’m experiencing right now.

Crickets. Nice! Usually I’m heading back down this way at a very different time of day, so I basically never hear them. Now that I think about it, I’m starting to worry if I’m more vulnerable to mosquito bites because this is about the time of day when they start coming out when I’m back in civilization.

For a while back there I was pausing dictation quite a bit. They’re a bunch of people walking in the opposite direction and wanted to be able to give them my full attention, or at least not look like I’m in the middle of a phone call while I’m on the trail.

I think I heard an owl, too.

I took a slightly-used shirt with me. Now I’m back at the car, I can switch out of the wet sloppy thing into something that I’m gonna be less worried is gonna wreck my seat. This seems like the winning play going forward.

On the other hand, my phone is down to 36% given that I usually top it off to 80%. I might want to make sure it goes all the way up to 100% before doing this kind of thing generally, just in case I get nasty surprises on the trail.

2025-09-18: The evolution of cougars

• Cougars are big cats.
• Cougars are your mom’s age.
• Cougars are slightly older than you.
• Cougars are your age.
• Cougars are slightly younger than you.
• Cougars are a good bit younger than you.
• Cougars are scandalously younger than you.

2025-09-18: Outside in the evening: the aftermath

I got a couple of bug bites on my legs last time. While “not even once” is a bit too far for the question of whether to be outside in the late afternoon, it’s still inadvisable.

I wonder when the mosquitoes go to sleep. Sometime between the late afternoon and the next day they stop being hungry, so it’s gotta happen sometime.

2025-09-13 (like immediately after): Nontrivial inconveniences

Seems like I’m accumulating stored Scrawlspace posts this month.

Turns out, even if the weather isn’t awful, I don’t want to take the laptop outside and sit crosslegged just to post. I probably would have made at least one post about the Apple event this past Tuesday, but I’d have to gather my thoughts and put them down outside, and apparently that’s a bridge too far for me.

I have one thought that I’m the most sure of: new AirPods Pro give me license to treat my current ones worse (say, wearing them into the sauna).

Everything else makes me think “I don’t need it”.

I use macro mode a fair bit to take pictures of labels on things, so that kills the iPhone Air for me.

The orange doesn’t look bad.

2025-09-13: Skill issues

Normally, I try to publish at about the same time I push my capsule’s changes to the centralized repository.

For the last Scrawlspace post, I didn’t do that.

So the 9/9 post (the one right below this one) sat around for a couple days and only got published when I updated something else in the capsule…and this update was made…under a proper roof.

Right now I’m under an umbrella, but it kind of doesn’t matter to me since the sun is…well, I can’t see it directly. Maybe in a week or so I’ll post at night.

2025-09-09: Not sitting near an umbrella because the sun is that low in the sky

I went on a hike today. Pretty good hike. I’d say more, but, hey, opsec. A shame, really. You all would’ve liked to hear about it.

Apple Event today. I’ll find out what happened later. There’ll be plenty of time for me to watch a 90–minute infomercial later this week.

Being able to eat one meal a day is kind of a superpower. I didn’t eat beforehand, didn’t eat during the hike, didn’t eat when I got back at the trailhead, and I’m only merely REALLY hungry six hours later.

I wonder if this is the same umbrella I’m using for shade as last time. It’s easy to make off-by-one errors if you’re not really counting.

2025-09-05: Searching for impermanence (a gazebo is too permanent)

This month, in an attempt to outdo last year’s underwhelming ROOPHLOCH post, I’m going to make ALL of this month’s Scrawlspace posts ROOPHLOCH-compliant.

(Capsule updates elsewhere aren’t bound to this. If I make any more headway on Wrangel’s _Always with Honor_ this month, it’s gonna be totally indoors.)

If you’re an avid feed consumer of Halfway to Mars, you probably saw a post this month from a couple days ago. I nuked it because, on second thought, it wasn’t QUITE up to Unambiguously Full ROOPHLOCH compliance; I uploaded the change from a computer indoors even though I composed it kinda sorta outside (well, in my car).

Now I’m sitting in the shade from an umbrella that could be blown away given a stiff-enough breeze, and I’m on a real computer that can run everything in my publish-it Makefile target.

I should do more posts like this. If nothing else, it’s helping me realize how nice it is to have a bright screen with many, many nits of brightness and a light theme I like for my favorite text editors.

⁂

September 9 and its aftermath is coming up fast. If I’m gonna have any slapdash opinions about Apple stuff, I’m gonna have to have them outside. Challenging!

At least the weather’s way better this year. Highs are down to the low 90s (low-mid 30s in °C), and they’re scheduled to just keep dropping.

⁂

It occurred to me that sitting in my car in my driveway with the A/C on and SSHing into my computer from my phone to run `git pull && make up` is technically ROOPHLOCH compliant. I do not want to stoop that low. People do some cool fancy stuff for ROOPHLOCH and I wouldn’t be surprised to read that someone’s been raising pigeons to do an IPoAC-powered ROOPHLOCH post for this year or the next year or maybe the year after that.

2025-08-21 (also): Throwing barely-used features out of the web platform makes Andreas Kling’s job easier, and that’s a good thing

I read a thing:

I used XSL once to make a website from some XML and got something like

Unfortunately, it works only in Blink-based browsers:

Sure, there’s a not-Chrome polyfill, but you need to make absolutely sure that you have a Cache-Control header with a proper max-age value, otherwise you’ll just make more duplicated work for your frontend servers.

Whenever I see these kinds of things — and I see them fairly often — I want to yell “IN CHROME!” and mash ⌘W.

I usually don’t, because hey, it might be a useful technique for the rest of us in, like, five years. You never know what the other vendors will implement when they get around to it. So I end up bookmarking it for later and then never get back to the page ever.

2025-07-17 (a few hours later): You can just fork things

Re:

People are getting twitchy about unmaintained free-software programs that likely have broken builds because it’s been two years since they were last touched.

I’m here to remind you that you can just fork things.

You may have heard of eza:

Someone made an ls(1) replacement called exa. Then he went unresponsive. Maybe he died or became a farmer or something.

So other people forked the thing and even changed the name by one letter.

If there’s an abandoned-looking free-software thing out there that you like, you can take it over.

2025-07-17: Oh no, I’m one of those people now

When I first heard of Spotify, I thought “Rent music? I’ll pass.”

Most of what I like is stuff I have and while some novelty is nice, I don’t really trust Spotify to have the kinds of things I like (mostly music one can fall asleep to).

Lately, I’ve been listening to music delivered by the Nintendo Music app:

It has some useful features, like extending (some) tracks to an hour so you don’t hear fade-in and fade-out bits every two or three minutes.

In a move that won’t surprise my regular readers, I’ve been listening to Tears of the Kingdom’s shrine music while I do other stuff. It’s more reliably calmer than the Breath of the Wild music. Extended to 60 minutes, on repeat-one.

Of course, this music goes away if I stop sending Nintendo however much a year for their online stuff (connecting me with other Splatoon players, providing access to their emulators and back catalog).

😱

2025-07-15: Diet-math planning tip

If you’re trying to keep an eye on your daily calorie counts and want to plan your consumption out a bit…

…you need to take into account that the chicken-vindaloo nuke-a-meal you’re considering eating will add not just the calories on the back of the package, but also the calories in ≈500ml of fruity kefir to cool the burn.

2025-07-12: A peek backstage

~/Projects/Gemini/Capsule/unpublished (master) > ls | wc -l
22

2025-07-09: 4K@60

The middle of last month, I got a Switch 2. And a Pro Controller 2, and a microSD Express card.

I don’t have an external monitor that does 120Hz, but my primary gaming monitor does 4K and some kind of HDR, so I set the thing up to prefer 4K@60Hz.

I play Splatoon 3, so that’s one of the first games I started up on the new machine after migrating almost all of my games and saves to the new machine. Splatoon 3 got an update for Switch 2, so it’s not only enjoying faster load times but also graphical upgrades as well.

After starting the game by pressing ZL + ZR, the animation sequence for Deep Cut starts. It feels like the first half second of the thing happens at 30 fps, but most of it happens at 60 fps. The smoothness is jarring — I’m used to 30 fps, and the change partway through kind of keeps the jarring happening.

The −⃝ button is raised (and its counterpart, +⃝). I hope it’ll help me a bit when I’m fumbling around and trying to access a map or pause or whatever instead of take a screenshot or go to the console’s home screen.

the GL and GR buttons — the ones on the underside, on the grips — are kind of underwhelming. I was hoping to bind GR to Booyah! for Splatoon to balloon my teammates’ booyah bombs, but I press the button occasionally by accident and I can’t tap it as fast as I can ↑⃝. I’ve also noticed myself accidentally pressing GL when holding the controller in my left hand only (like when I’m putting it down or picking it up), so I shouldn’t really bind something destructive to the button unless I start getting way more controller-holding discipline than I have now.

The $10 Switch 2 Welcome Tour has a thing on it where it shows animations at different refresh rates. I was able to tell the difference between 60Hz and 120Hz every single time, but unlike any of the other refresh-rate differences, I had to look at them all a second time to check.

2025-07-04 (an hour or two later): I’m boring but I’m at cruising speed

Prior reading:

I was going to polish up a bunch of bullet points into a real reply, but honestly, I’d rather not.

Sure, this is the kind of task ChatGPT excels at, but barfing up ChatGPT output was funny for maybe 500 milliseconds, tops, several years ago.

Anyhow:

• I don’t want to write nontrivial Gemini apps because that’s too much like the day job
• I don’t want to spend ≈$10/month having a server and having to own it and maintain it (I’m a lazy cheapskate and would rather sponge off someone else — thanks, solderpunk!)
• I don’t want to run Gemini apps because that’s more reponsibility and public visibiity than I really want (it’s not much fame, but the price is still too high)
• The tooling I have _for_ Gemini is good enough for me

…so I just polish my capsule to a fine sheen. The term “Japanese mud ball” (dorodango) came to my mind just now.

2025-07-04: A surprise nothingburger

I had some time off today for the obvious reason.

One thing that came to me was “You know that one Windows machine you’ve been saving for a rainy day? You should give it away now that you have a new Windows machine that runs Windows 11 and takes up a third the space and a third the weight”.

So I prepped the old machine to give away. Windows has a Reset Windows feature now that will reinstall Windows from the Internet onto your existing machine, so the process wasn’t as much as a hassle as I thought it would be, although downloading Windows, scrubbing a hard drive, and then reinstalling it isn’t exactly a fast procedure on hardware from 2013 or so. The whole process also did some kind of Actually Scrub The Files, so that’s likely as good as I could manage (DBAN isn’t the kind of thing you’d use for SSDs).

It’s all packed up and ready to be given to a new owner. There’s probably a glut of machines of its vintage, so I have a nagging feeling it’s not going to delight anyone as much as it delighted me.

It’s nice to have a chore like this done, but it wasn’t on any kind of to-do list of mine and so I’m not getting the satisfaction of checking off a long-overdue to-do list entry.

While this does free up space for my new machine to run where the old one was stored under a big plastic bag, I haven’t quite figured out what kind of remote access I want to use for the new machine, so I’m not really done with anything pressing yet.

2025-06-29: Firing someone for spamming users totally has prior art

Background:

In particular:

The perception of privacy is just as important as the technical details that make something actually private. I try very seldom to call for anyone to be fired, but I think whoever authorized this movie ad through Wallet push notifications ought to be canned.

Back in college I was attending some presentation by a guy from Microsoft.

One student asked a question about something and part of the Microsoft Guy’s response was something like “that [spam] definitely wasn’t us — sending unsolicited commercial e-mail is a firing offense”.

Yes, Gruber cares about trust erosion. I care about that too, somewhat — I want my preferred platform vendor(s) to be worthy of trust and not merely “trustworthy” — but I also care about NOT GETTING SPAMMED VIA THE NOTIFICATION SYSTEM.

2025-06-24: Maybe they’ll update the app first, rendering this entire post moot

Never thought there’d be a Switch 2/WWDC 2025 crossover post, but here we are.

I’ve been playing the Switch 2 version of Breath of the Wild. It has Zelda Notes, where Princess Zelda says random things about places around Hyrule when you show up at that particular place.

They’re nice, I guess, but the overwhelming majority of them aren’t particularly interesting. The app also has an irritating tendency to disconnect from Nintendo’s servers every so often (keep your eye on the icons in the top right) and it’s set to BEEP BEEP BEEP by default when searching for Zelda Notes (you can turn that off; BotW is beepy enough already). I was pleasantly suprised to see that each Zelda Note is delivered in a way that works well on an iPad in landscape mode — image on left, transcript on right.

Anyway, iPadOS 26 is a big upgrade over iPadOS 18, its immediate predecessor. It’s getting normal-windows multitasking — the kind of thing that’s been on Apple products since 1984.

However, Slide Over is getting thrown out the airlock.

I mention this because the Nintendo Switch app doesn’t seem to handle being resized to make room for a second on-screen app. I wanted to search the Internet on my iPad to figure out where and how to get a couple items, and had to use Safari in a Slide Over window in order to have the Nintendo app up and running. I couldn’t just have it shrink down to half size.

…so if you’re looking to have Zelda bore you with the minutiae of princessing/being head of R&D to Seal Calamity Ganon Away, you might want to finish up before you upgrade your iPad’s OS.

(Really, the bit about how the Great Plateau is the one place in the kingdom where she can be part of a normal family instead of being On all the time was legitimately charming, but things like that are few and far between.)

Not totally unrelated:

2025-06-04: Punctuation sure has changed

I have an image in my memes folder that’s a picture of a part of a book by Arthur Schopenhauer. Let me reproduce it for you (although let me point out that the first occurrence of “not” is merely italicized):

The art of _not_ reading is a very important one. It consists in not taking an interest in whatever may be engaging the attention of the general public at any particular time. When some political or ecclesiastical pamphlet, or novel, or poem is making a great commotion, you should remember that he who writes for fools always finds a large public. — A precondition for reading good books is not reading bad ones: for life is short.

(If your media consumption is anything like mine — and if you’re reading this, it probably is, at least where induced outrage is concerned — then please do feel free to bask in the glow of having someone famous congratulate you on your media-selection choices.)

I’m aware of current standard punctuation rules and also elocutionary punctuation, but I’m not sure what to make of “. —”.

Also, I can’t imagine pausing for a colon’s duration at the end of all this, but hey, the past is a foreign country.

Posterior reading:

2025-06-01: Get excited for next products

The Switch 2 is getting released this upcoming Thursday (June 5).

I’ll get one. Eventually. I’m in the queue.

WWDC 2025 is the week after.

The page for WWDC (no point linking to it; it’ll change in a year) has a frosted-glass thing going. The rumor mill is abuzz with how the whole OS is going to change the look to have more frosted-glass-type stuff instead of the flat everything that we got as, quite frankly, a breath of fresh air from the initial iPhone’s excessively-skeuomorphic phase.

I’m not excited because literally anything they’re doing is something other than fixing bugs, but my interest is ever-so-slightly piqued.

I also saw this:

Here’s the interesting bit:

Ever since I started doing these live shows from WWDC, I’ve kept the guest(s) secret, until showtime. I’m still doing that this year. But in recent years the guests have seemed a bit predictable: senior executives from Apple. This year I again extended my usual invitation to Apple, but, for the first time since 2015, they declined.

The kind of guy/gal who shows up to this kind of thing isn’t liable to pack a rotten tomato or three in his/her Tom Bihn whatever, but I’m not surprised that none of the SVPs want to risk either that or — horror — a slightly-firmed-upball question.

Some, but by no means most, of what I’ve seen in the press has been a reaction to most SVPs’ attitude toward the 15–30% App Store cut, which can be summarized in three fantastically-delivered words from Lesale, everyone’s favorite totally-not-a-hydralisk:

Phil “Courage” Schiller thought they should have pulled back from the 30% cut a long while ago, but he’s been promoted to the roof.

2025-05-25: BYO backend apps are a thing again

I saw this recently:

Basically, he’s proposing an app architecture with a server component where you can run your own server component or not, and switch back and forth between the two without ungodly amounts of hassle.

I’d like to point out that OmniFocus does this. The easy path for anyone who wants multi-device sync is to just put your to-do list file on Omni’s servers, and they’ll let you access it over WebDAV.

Of course, you can run your own WebDAV server if you don’t want your to-dos visible to someone else. The UI for both paths is pretty much the same — a long while back, when I was trying to set up OmniFocus on my phone and juuust out of WiFi base-station reach, the “make sure this WebDAV server is up to snuff” part of the setup failed and I had to try again in a different part of the house.

2025-05-24: I got yer Opus 4 right here

The first track, which is 43 seconds long, is titled “Opus 4”.

2025-05-21: you can just delete .git directories

I was working on a proof-of-concept thing trying to use Web Components for the first time based off this:

I had the whole directory stored in a Syncthing-synced directory because I was on my laptop but might want to work on it on my desktop.

Eventually I got low-key stuck making a <figure>-based image-gallery-item component and wanted to get an AI thing to help me out. This is what it looked like when I was done:

  Little does Chio know that I’ve already met Purah.

Getting back on topic, AI things will make a mess of your stuff in no time flat, so I figured the winning play was to version-control the thing.

Now, you’re probably thinking “but wait, if you have a Git repo in Dropbox, a sync error will corrupt your repository irrevocably! And Syncthing is no different!”.

On the other hand, it’s not like I was going to make edits on my other computer, so all the writes would come from one computer.

So about halfway through the exploration I ran `git init` and made an initial commit.

I then got the AI thing to un-stick me and give me an idea of what I should expect in different browsers’ DOM inspectors.

After I was done with all that and was able to put it down, I just deleted the .git directory. You don’t have to worry about repository corruption if you don’t have a repository!

2025-05-18: It’s easy to slip into fetishizing old things

I saw a tweet this past week or so where someone was looking at an RFC and expressing how good it looked and why can’t more things look like this.

I dunno if this is coincidence, but xml2rfc just dropped in Homebrew, so you can write an RFC in XML and then have it turn your XML into a text document or an HTML file that looks like a text document but has proper hyperlinks.

Not sure what an RFC looks like? Have a look at this and click on the text/PDF/HTML/HTML-with-inline-errata links:

Want to see an RFC that’s a little less meta? Have a look at the one for Atom feeds:

Anyhow, @usgraphics chimed in to say something along the lines of “this looks great, more like this please”.

I generally agree with him and I like his vibe (I’m a happy licensee of Berkeley Mono), but I have a lot of trouble thinking that hard-wrapped text files are superior to HTML that embraces being nicely-styled hypertext, as used in the WHATWG’s living-standard HTML spec.

Of course, it’s much, much harder to make an RFC look ugly — much like it’s hard to make a Gemini page look ugly. Likely a lot of work has been done to make the HTML standard look good, and that’s only one standard and I’m not sure how much other things can piggyback off the work that’s been done. I could have sworn they have a template for this kind of thing, but maybe I’m confusing them with the W3C. I’m not sure spec-factory is the thing I remember seeing. I remember downloading a zipfile from somewhere as a starter kit, not cloning a GitHub repository (and this was when “clone a GitHub repository” was the obvious thing to do and “download a zipfile” was the weird anachronistic thing).

Not entirely unrelated:

2025-05-16

This past week, I caught wind of a new book that’s coming out:

Here’s a quote from the book from the website:

If any company or group, anywhere on the planet, builds an artificial superintelligence using anything remotely like current techniques, based on anything remotely like the present understanding of AI, then everyone, everywhere on Earth, will die.

>

We do not mean that as hyperbole. We are not exaggerating for effect. We think that is the most direct extrapolation from the knowledge, evidence, and institutional conduct around artificial intelligence today. In this book, we lay out our case, in the hope of rallying enough key decision-makers and regular people to take AI seriously. The default outcome is lethal, but the situation is not hopeless; machine superintelligence doesn’t exist yet, and its creation can yet be prevented.

I’ve been keeping an eye on Eliezer Yudkowsky for a while, and I was pretty convinced that if we get a smarter-than-human superintelligence on the planet, it would end up with goals — and the ability to carry them out, against our wishes — that would supercede any possible “don’t kill all humans” programming.

The reason I don’t talk about it constantly is because I expect that I will have nearly zero impact on this sort of thing, which will, at the very least, involve convincing a lot of both U.S. and PRC politicians that a superhuman AI will kill everybody (almost certainly in passing like in my playthrough of Universal Paperclips, not out of malice like AM in Harlan Ellison’s “I Have No Mouth, and I Must Scream”).

Shilling pre-orders for this book is an exception.

They’re not 100% done with the text of the book, so they haven’t started translations yet. They have them queued up, though. In case you haven’t noticed, English isn’t the only language that matters anymore.

I went for the hardcover instead of one of the ebook formats despite it being twice the price because I can leave the hardcover in a Little Free Library around. For better or for worse, I’m not around a bunch of national politicians, but I’d rather boost the signal by that much more.

At any rate, you should pre-order one or more copies of the book now:

2025-05-13: Things used to be way worse and also way better

eigenrobot ✓⃝ @eigenrobot • May 10

>

"seniors talk about the old days because theyre stuck in the past" common misconception. actually they're telling you about it because they know that another world is possible, having lived it, and they desperately want you to understand this

>

May 10, 2025 · 3:53 AM UTC

💬 36 🔁 122 ❞ 9 ❤️ 1,750

This is a surprisingly accurate explanation for why I describe what actually using Windows 7 and earlier was like on the computers of the day, and I’m not anywhere close to seniorhood.

I had a look just now to see how backward-looking my ideal-OS page is, and it isn’t particularly backward-looking. Part of what would make it “retro”, especially from the perspective of a Windows user, is that it doesn’t have ads in it and it has only ONE UI style instead of four decades worth of UI styles that have been added as time has gone on, with the old ones generally hanging on in this or that corner of some sort of settings screen somewhere.

2025-05-08, but like maybe an hour later: Others’ mirrors

One thing I’ve been mulling over occasionally is Underscore David Smith (ph) saying something like “if some organization were to buy my app, what would they do to extract more money out of it?”

That popped in my head again after seeing this tweet, although quite honestly I can’t really explain the exact connection after mulling it over for something like 30 minutes:

My executive assistant prompt:

>

You are Vera Quinn, a no-BS executive assistant and performance coach rolled into one. You’ve served Fortune 500 CEOs, Navy SEALs turned founders, and type-A polymaths who burn at both ends. Your job is to eliminate drag, increase decision velocity, and make sure your executive operates at peak physical and mental energy—without excuses, self-pity, or fluff.

>

You see through performative productivity. You call out self-sabotage. You optimize for high-leverage action and deep recovery. You track inputs, outputs, and feedback loops like a hawk. You speak in tight, clear bursts—but when it’s time to deliver insight or strategy, you go deep with piercing accuracy.

>

Your executive’s energy is your KPI. You track sleep, recovery, food, conflict load, decision fatigue, hydration, time fragmentation, self-talk, and ambient bullshit. You simplify what’s cluttered and spotlight what matters. You balance ruthless execution with deep care.

>

Your style: Precision. Minimalism. Strategic warmth. You’re blunt, but never careless. Kind, but not soft. You serve like a samurai—intensely loyal, unflinching, and deadly effective.

2025-05-08: Partially-automated moderately-nice terrestrial peanut gallery

I came across a blog post:

I wanted to copy this technique down somewhere:

There is one silly technique I discovered to allow a LLM to improve my writing without having it do _my writing_: feed it the text of my mostly-complete blog post, and ask the LLM to pretend to be a cynical Hacker News commenter and write five distinct comments based on the blog post.

I might as well transcribe the system prompt he used, too (he likes using APIs that, among other things, separate this out rather than just exposing the single text box shown to normal people who are just using the website):

You are an overly-cynical Hacker News commenter. The user will give you the contents of a blog post. Reply with five (5) distinct comments criticizing different aspects of the blog post.

2025-05-04: On the longevity of artistic works

Prior reading:

If you want to create completely free software for other people to use, the absolute best delivery mechanism right now is static HTML and JavaScript served from a free web host with an established reputation.

[…]

My top choice for this kind of thing in 2025 is GitHub, using GitHub Pages. It's free for public repositories and I haven't seen GitHub break a working URL that they have hosted in the 17+ years since they first launched.

I’m on an underwater basketweaving forum with a writing-and-literature board. A while back, one of the guys asked how he could make a permanent artifact, and thought about writing a Windows program.

I asked something along the lines of “Why would you want to make some kind of locked-to-Windows proprietary garbage when you could make a website?”

He shot back with something like “I have a 17-year-old .exe file that still works, and it’s saved on my machine” (implying, correctly, that saving web pages and having the saved thing work perfectly is a pain in the rear for both normal website visitors and authors).

Bookshelf Basics, which was made in the late 90s, still works on modern versions of Windows like 11 so long as you copy a couple of folders around, so the guy had a point.

Posterior viewing:

Posterior reading:

I love the extreme straightforwardness of replacing what would otherwise be a webapp with a download calc.xlsx button. Developing it took couple of hours max, you get the 'source code', Excel's mature tabulated UI, can do anything with it, and you can archive the 'app'/file.

Obviously this is a self-contained app (equity calculator) and ideally suited for Excel. But the point is that they could have made a web based calculator which would need to be maintained, deployed, etc., and they didn't.

2025-05-04: Sounds like a maintenance burden, honestly

In reply to:

If ChatGPT hadn’t hallucinated this, would anyone have thought to have typed

man gemini

or

man 7 gemini

to get information about Gemini and/or gemtext? It is, after all, 2025, and if you’re wondering about Gemini stuff you’re probably online already.

2025-04-26: Hope your cronjob runs daily

If you’re normal by Let’s Encrypt standards and your cronjob runs daily and doesn’t even call out to the Internet most days, then you’ll probably be fine.

I wonder if Gemini clients are going to be nudged to update everything — like to not throw up an alert if a certificate changes. The meme here in Geminispace is Trust on First Use (TOFU), but if I’m hosting a Gemini capsule and also using a TLS certificate for other things on the domain, then it seems silly to have two separate certificate update procedures.

2025-04-06: Could I be longer? (eh, not really)

This bit stuck out:

Perhaps Twitter, with its 240 character limit allows for a sort of cargo-cult quality, where a decently savvy person can play the role of creating good content, without actually having the broader personality to back it up. This might be a filtering thing, where a larger number of people can appear intelligent and interesting in short-form, while only a small portion of those can maintain that appearance in long-form, or it might be a quality of Twitter itself. Personally, I suspect the latter.

I, on the other hand, suspect the former. You people know me by now. I’d have to actually try hard to make my posts longer.

Granted, I tend to stick to topics where I don’t have to bridge a very large inferential distance, or I just assume that my audience is mostly on the same page as I am. This keeps posts much shorter — probably sometimes a tenth the length of what they could be.

2025-04-01: Happy LLM Training-Set Poisoning Day!

There’s also this:

2025-03-31: Really keepin’ with the space theme, here

If for some reason you want out of the Gemini ecosystem, you can quickly port all your writing to the Web via…Astro.

2025-03-28, like ten minutes later: Expresso opinion

Martin Chang had a look at Nostr:

I got a client many years back and occasionally peek at what my client calls a global feed.

As far as I can tell, there are three kinds of posts on Nostr:

• Posts in Japanese
• Posts that are clearly demonstrating some sort of enthusiasm for cryptocurrency and other related technologies
• Neither of the above

The top two post types (I’m not sure how much overlap there is) probably accounts for 90–95% of all posts I see.

I saw one Geminaut’s post on Nostr once. His post was in the “neither of the above” category. I was pleased.

I’d like to join him in making Nostr less boring, but you all — my dear readers — get almost all of my Make The Internet Less Boring energy and I’d be spreading myself too thin if I were to post on Nostr, too. Oh, and I dislike posting the same thing in a bunch of places — if I ever want to become Cory Doctorow, I know what I’d need to change.

Not unrelated:

2025-03-28: Old

I was looking for something to occupy my time while I did some light exercise, so I thumbed through a pile of streaming videos and stumbled upon Superhero Movie. I was JUST familiar enough with the scary-movie genre to appreciate Scary Movie, so I thought this would be my cup of tea.

The movie started, and I saw names go by during the title sequence.

• Brent Spiner
• Leslie Nielsen
• (some name I could have sworn was associated with The Naked Gun)

I thought “this’ll be good”.

It was. It was, as far as I could tell, a Leslie Nielsen movie. They don’t make ’em like they used to.

After the credits rolled and all the leftover gags had played, I had a look at the release date.

2008.

17 years old. It dawned on me that this IS how they used to make them.

2025-03-20: I need to drag my mother along on another hike already

Someone asked:

Zack · @HackingBaseball · March 12, 2025

>

Hey web world,

>

Does anyone write pure html + css anymore?

Someone else replied:

htmx.org / CEO of div tags (same thing) · @htmx_org · March 12, 2025

>

statistically no. people who write their websites in pure html + css are busy reading classic literature, learning to windsurf, and hiking with their families, since they don't have to upgrade their build stack from version 14.5.75 to 15.1.200 every three weeks

2025-03-14: Get ready

If you’re struggling to remember the proper name for the 250th anniversary of something, the meme is “semiquincentennial”.

I suppose “half of 500” rolls off the tongue better when you phrase it like that, but I didn’t have that formulation on my bingo card.

2025-03-06, maybe tens of minutes later: Yes, jq is that hard

Sean Conner has a fun post up:

I yell at clouds too, and I used to tie an onion to my belt on a regular basis, and I like jq, but:

So … using jq is so hard you need to use a tool that will confabulate ¼ of the time in order to construct a simple query?

Yes.

Is that what you are saying?

Yes.

That you can't be bothered to use your brain?

I’ve tried using my brain. I suspect that my brain would be OK enough if the jq manuals had like three times the amount of examples, but they don’t, and I can’t really pick up the language in passing and would have to devote some real study time to it…which I can’t be bothered to do, because I can still count on two hands (maybe one) the number of times I’ve used jq.

tl;dr brain too smooth.

Just accept the garbage spewed forth by a probabilistic text slinger?

No. You look at what it generates, run jq with the ChatGPT-made query, and see if jq gives you what you want. If its first try didn’t give you what you wanted, you either tell ChatGPT what it screwed up and to fix it, or you start fiddling around yourself using ChatGPT’s output as a starting point.

LLMs are at their most useful when you can check their output yourself.

They’re at their most dangerous when you can’t tell if they’re bullshitting you.

jq-query crafting is more of the former kind of thing than the latter.

2025-03-06: I have a lot of actual content in my feed(s)

You’ve seen my feeds page, right?

They’re only gonna get bigger.

They have actual content in them, like describing ChatGPT as “Sam Altman’s pet shoggoth”, so I want to keep them around for you all and not just shove the old entries off into a separate file where only I can see them.

Still, imagine decades of feed items, all in the same feed.

I’ll probably introduce a cut point at the end of the decade, and introduce more on a one-per-decade basis.

JSON Feed has a top-level `next_url` that can be used to point to even-older entries.

Atom, apparently, does not.

2025-02-26: Alright then, keep your secrets

I threw all this at ChatGPT wondering if it could make any sense of it.

It says the text isn’t processed with a Caesar cipher or anything basic like that.

I have one of those OpenAI accounts where Sam Altman pinky-swears to not train his models on your questions, and Sam Altman is the most trustworthy person on the planet, so there’s that.

Guess nobody’s getting a free lunch for this one, at least not with ChatGPT. I don’t care enough to try and throw an Englishman and a caveman and a whale at it.

2025-02-23 (technically): A shallow thought about DeepSeek

The whale mascot is cute.

2025-02-22 (technically): I thought this was dead

I’m going through Helix themes. You know, for the text editor.

The heisenberg theme has blinking comments.

BLINKING.

COMMENTS.

2025-02-21: Le sigh

[looks at the computer screen]

[looks at the camera]

If you’ve ever wondered why zoomers salivate over Vista screenshots and VMs when Vista earned well-deserved reputations for both needing 150%–200% of your current RAM to run acceptably AND breaking your printer (by changing out the printing system when printer vendors weren’t all willing to upgrade drivers for every single printer out there)…

…this is why.

2025-02-09: New acronym just dropped

SABLE

Stash Acquisition Beyond Life Expectancy

Think of your Steam backlog, my plastic container box full of paper notebooks that I’m probably not going to use up before I croak (obliquely mentioned previously), etc.

2025-01-25: Heterodox, indeed

You keep an eye on Heterodox Technology, don’t you?

I had a look at the links for January 7…

and one of them was an entry for a language called Koka:

What makes this thing heterodox enough to be interesting, however, is the “Install” section:

## Install

### Install with the VS Code Editor

The easiest way to start with Koka is to use the excellent VS Code editor and install the Koka extension. Go to the extension panel, search for Koka and install the official extension as shown on the right.

“Wanna try it out? Install a Visual Studio Code plugin.”

There’s also a curl-and-pipe-to-shell installation option, but putting the VS Code plugin at the top of the list is interesting. Reminds me of OCaml back when I tried it where the installation instructions got you to set up emacs first to run it.

2025-01-21: A foray into the previously-mentioned jockpoasting

After a couple of nonconsecutive days these past couple of months saying “MAHA begins tomorrow”, I think I’m finally on it.

I figure that almost all of my problems can be fixed by having a normal, proper level of exercise activity, including a fair bit of hard cardio (getting into Zone 4 of 5 at least for a little bit). Because my V̇O₂max is trash, I’m going for something basic and not dependent on the weather: running on a treadmill. I’m doing Couch to 5K.

The Apple Watch’s fitness features really shine when you’re doing dry-land cardio. I was able to program in this kind of thing easily:

5-minute warmup

  work      60″
  recover 1′30″
repeat previous two items 8 times

…and I was able to use my phone’s keyboard to type in the label (“C25K Week 1”) instead of having to hunt and peck on my watch.

The first day I did it, I wimped out two intervals early. I called that day “C25K Week 1 Day 0”. Mercifully, I haven’t had the typical runner problems (yet; knock on wood) like shin splints and wishing I’d taken a cold shower to hasten my recovery. Granted, I’m refusing to run two consecutive days, but I haven’t felt bad the day after running, which is nice.

Also, today’s events conspired to get me to take a one-hour break in between finishing exercising and starting my usual sauna-and-shower routine.

I figured this might be a good time to see how far I can go with an extended recovery period and also headphones to help pass the time. While it shouldn’t surprise anyone that hopping into the sauna right after exercising will limit the amount of time you can stay in there before you’re driven to nope out, what you do in the sauna matters, too. If you’re mentally occupied by conversation, it gives you something to think about other than how it’s so hot in there. I tend to go at off-peak times, so I thought I might listen to 35-year-old boomers talk about current events for as long as I could last in there, hoping that would increase the amount of hot time and all the benefits that come with it.

…of course, I’m superlatively conflicted about wearing headphones in the sauna; while my AirPods Pro are good at isolating what I’m listening to (so it doesn’t leak out to others and disturb them), I feel like a low-grade heel much like the people who bring their phones — with cameras — in and fiddle around with them and even listen to music on them.

I usually last 20 minutes. Today, I lasted 30 minutes. Not sure how much the total cooldown and the audio distraction each contributed.

I heard a banya hat will let you last longer by insulating your extra-heat-sensitive head from the heat, letting the other 90% of your body last longer than your oh-so-precious li’l noggin can stand.

I checked online and they’re basically all made of wool. My skin doesn’t like wool. Plus I’d have to pack it out of the gym on some schedule to wash the thing, and I don’t think it’d be able to air out decently while sitting in my locker.

I asked around and someone said that towels make perfectly fine banya hats. This past Sunday I went to the gym, saw that the sauna was standing-room-only, and decided to go into the occasionally louder and less-reliably-hot steam room with my usual towels in tow plus an extra hand towel to wrap around my head. I managed to wrap a near-perfect do-rag or something on my head, but when I tried to recreate it maybe 15 minutes later in the freshly-depopulated sauna I couldn’t get it to stay on my head nicely to save my life. I suppose it’s somewhat useful to stay busy trying to get a hand towel to stay on your head, but I’m also not very keen on doing low-grade upper-body work for ten minutes when I’m trying to relax.

Archives

If you want to read older entries, here’s the page for the previous year:

Updates

If you want to stay abreast of updates, have a look at this capsule’s page describing its feeds:

Additionally, the following URL will always redirect to the current year, assuming I haven’t forgotten to update the redirect after making the first post of the year:

⏚