i've been working on a pebble watch face for a bit and i wanted to share it cos i'm proud of it! i'm at the hardest part now, time zones, but once i finish that i'm kinda home free.

any thoughts? there's a lot of parts i want to change once it's finished and i can give myself permission to increase scope.

I don't know what sort of feeling this thumbnail is giving me, but I really need to unsubscribe from Scott the Woz

For now, here's some links.

Thom Holwerda: Dark patterns killed my wife's Windows 11 installation

I'm seething with rage towards Microsoft and its utter incompetence and maliciousness. Let me, for once, not mince words here: Windows 11 is a travesty, a loose collection of dark patterns and incompetence, run by people who have zero interest in lovingly crafting an operating system they can be proud of. Windows has become a vessel for subscriptions and ads, and cannot reasonably be considered anything other than a massive pile of user-hostile dark patterns designed to extract data, ad time, and subscription money from its users.

A quote I have heard from somewhere--I do not remember the source--is "While 2026 will not be the year of the Linux desktop, the 2020s will be the decade of the Linux desktop." Linux has slowly been getting more and more "okay" and Windows has become more and more dogshit. One day we can only hope that the world will reach a point where there is a viable alternative to being treated like total garbage available for most people.

Grayson Davis: Underwater Basket Weaving

"The bride is a graduate of Prairie Dog Tech, where she received her degree in underwater basket weaving. Her other pursuits of study were familiar bird calls, Egyptian folk dancing and men."
--1952 April Fools' Day article announcing "Chlorophyll I.S. Green Marries Ima Violet Cloud in April 1 Rites" (The Rock Island Argus, 4/1/1952)

Davis' new blog "I Might as Well Explain the Joke" is truly great. An all-timer blog worth following. Add it to your RSS feed!

Chris Person: I'm Tired Of These Useless Jackasses Making The Computer Expensive

I was planning on doing this as an Aftermath piece, but then I looked at the prices of spinning disk drives, which are now twice what I paid for them--just like SSDs, and just like RAM, just like GPUs before them.

I'm sharing this because, like the author, I love the computer but hate the environment around the computer.

Bruno Dias: Against 'Metroidbrania': a Landscape of Knowledge Games

I'd put it forth that the term doesn't just sound dorky, and it's not just hard to explain (given that first you have to explain what a metroidvania is -- itself a fraught bit of terminology). I'd say it's all but useless, too.

This is possibly one of my favorite blog posts of 2025 and I cannot believe I haven't shared it yet.

Clyde Mandelin: The Many Translations of the "Lost Woods"

If you recognize this name, it's because he translated Mother 3. He's got a whole
bunch of blog posts that talk about video game translation, an this one just so happened to be the most recent one.

I don't think there's much to say actually. I think this card wants be run as a 1 or 2-of in a decent number of non-bad publicity focused decks, and that means it's a pretty good card. Lots of money y'all. If you want more opinions on bad publicity, please read Veronica's piece on this card and Editorial Division.

Before I talk about Let Them Dream, I need to talk about 5/3s. The job of the 5/3 agenda is to win you the game in 3 agendas, and decks that run 5/3s usually want to score out with any combination of two 2-pointers and one 3-pointer. But the 5/3 has one major vice: they kinda suck. They're so much harder to score out than 4/2s (or 3/2s!), and their payoff usually isn't even worth the work! This is especially true because corps vastly prefer to score out 5/3s as the game-winning agenda, so any "on score" text is irrelevant. If the corp got to choose, they would dedicate the entire power budget of a 5/3 towards hurting the Runner, and none of it on helping the corp. This is incredibly obvious when you see what 5/3s corps run, and why they run them. For example, Send a Message and Next Big Thing give the corp tempo for having an agenda scored or stolen, while SDS Drone Deployment and Méliès City Luxury Line hurt the runner's tempo in the worst case and can deny a steal in the best case, but the best thing a 5/3 could do is to guarantee the runner needs to steal more agendas than you need to score. This is what Global Food Initiative did. It did so well at this job that it was used in 99% of competitive decks when it was put on the restricted list 2 years after its release.

Let Them Dream may have the same incredibly powerful effect as Global Food Initiative but, as a 4/2 that costs one neutral influence, it’s having to compete for deck slots with 3/2s that are easier to score and strong 4/2s that have a bigger impact on the board. Does this mean it’s bad actually? Hell no! Being worth only 1 point on steal is still too powerful. I can see it being used in maybe some slower glacier decks, and definitely in some decks that only run 1 and 2 point agendas (Thule), but I don’t see Let Them Dream being anywhere near as dominant as its predecessor.

Scoring PD is still back baybee! Ikawah Project has been a staple card in scoring PD for about as long as I can find decks of it. (Here's examples from 2021, 2023, and 2024. I have no idea what was going on in 2022.) The only card that's as much of a staple in these decks as Ikawah Project is Global Food Initiative (which might be relevant to my next card review). This might be bringing back enough old tools for scoring PD to make it a top deck in the Vantage Point meta. I know that I'm certainly gonna be playing it.

Hey y'all. This is gonna be a short one. My post about Arborium ended up doing severals. By this I mean it was briefly the number 1 post on Lobste.rs, which is an invite-only diet orange site. Thankfully, I've only received a small amount of harassment from this. Anyways Netrunner card reviews coming soon.

Last weekend while scrolling through some quieter feeds in my RSS reader, I came across an article which might be the something better I was looking for.

fasterthanli.me Dec 13, 2025

Introducing arborium, a tree-sitter distribution

About two weeks ago I entered a discussion with the docs.rs team about, basically, why we have to look at this: When we could be looking at this: And of course, as always, there are reasons why thi...

Arborium is a high-performance syntax highlighting tool powered by tree-sitter created by Amos Wenger. Amos (also known as fasterthanlime) is a long-time open source developer who I greatly respected. I learned Rust from his Advent of Code article series, and I appreciate his commitment to correctness and speed. I was especially excited to learn about Arborium because it was designed to work on the web using Javascript. I could use this!

Trapped in the Arborium

I started writing code to integrate it into my site immediately and had a plugin ready within an hour or so. Here’s the relevant code (edited for clarity):

codeElements.forEach(async (element: Element) => {
	const programmingLanguage = detectLanguage(element);
	// highlight is the all-in-one function from arborium that does all the real work
	const highlightedText = await highlight(programmingLanguage, element.textContent);
	element.innerHTML = highlightedText;
});

I tried building the site and...

Ouch... It didn’t work, but this is a simple problem. window is a global object that can be accessed anywhere in the browser because only the browser has a window. It doesn’t exist if you’re running Javascript code outside of the browser like with Deno.⁵ And if you try to use window then you’re advised to replace it.

I read through Arborium’s Javascript source code and found out that, as part of setup, highlight directly calls some code that tries to access window. No worries, there’s a more direct way of highlighting code that circumvents highlight’s setup phase. I’ll do that instead

codeElements.forEach(async (element: Element) => {
	const programmingLanguage = detectLanguage(element);
	// const highlighted = await highlight(lang, element.textContent);
	
	// loadGrammar tells Arborium to load the language you want to use.
	const languageGrammar = await loadGrammar(programmingLanguage);
	const highlightedText = languageGrammar.highlight(element.textContent);
	element.innerHTML = highlightedText;
});

And then I build my site again, and...

Ough... It didn’t work. I looked into it and this might be a Deno issue, which is less well-supported than the more commonly used Node. Arborium also appears to be doing some hacky dynamic code importing that isn’t playing nice with the way Deno expects imports to work. I haven’t tested this in Node, so maybe it works better there. Regardless, I’ve read enough of the source code to have learned that Arborium has some completely undocumented⁶ configuration options that would let me replace its dynamic code importing code with my own, so I guess I can do that.

codeElements.forEach(async (element: Element) => {
	const programmingLanguage = detectLanguage(element);
	
	// loadGrammar tells Arborium to load the language you want to use.
	const languageGrammar = await loadGrammar(programmingLanguage, {
		resolveJs: ({ programmingLanguage}: ResolveArgs) => (
			// Deno-specific dynamic imports to make the dang thing work 
			import(`npm:@arborium/${programmingLanguage}`)
		)
	});
	
	const highlightedText = languageGrammar.highlight(element.textContent);
	element.innerHTML = highlightedText;
});

That should be everything. In theory, I’ve worked around all these tiny little bugs and now I’m on the easy street of straightforward code that just works.

It should work. I should be done.

I’m back at the first bug.

So here I’m stuck. I try to work around the code that uses window and I run into issues with imports, then I try to work around imports and I run back into issues with window. There’s only one option left, and that’s to file some issues on the GitHub code repository for Arborium and wait for Amos to fix them.

Filing issues

I file two issues, one at Sunday night and the other on Monday morning. About an hour after the second issue, Amos responds saying he’s got a completed pull request⁷ that fixes all the problems I was running into. This seems like a quick turnaround for some environment-specific bugs, but he’s also a pretty good developer who’s been working on this for much longer than I have. Perhaps his experience meant he could quickly clock what the problem was and fix it, but I then open the PR and find my greatest fears realized.

You see, this whole time I’ve been misleading you. While I was trying to debug all these simple yet awkward bugs, a sort of dread was building up in the back of my mind. I was starting to get this feeling that Arborium might be almost completely written by AI.

To demonstrate what I mean by getting a feeling, let’s look at the Arborium website. It’s a nice-looking website, but if you look a little closer you’ll notice some inconsistencies: Arborium’s copy has a few, but not many, AI tells;⁸ some incorrect or outdated links; all code blocks except the “hero image” code block lack highlighting, which means they’re not formatted in a way that Arborium would recognize (something I would avoid if I were trying to show off my syntax highlighting tool); and, most importantly, there’s a complete lack of documentation. You’re given steps on how to use Arborium in simple use cases, but you’re completely fucked the moment you step outside the golden path, leaving you with no recourse except to read the source code. As you saw earlier, this is what happened to me.

This all added up to this strong AI “vibe”, but I wasn’t looking for it at the time. I had heard some time ago that Amos was dabbling with AI, so I didn’t come in expecting pristine, exclusively human-written code. But, I also did not expect the rot to go this deep. I suspect that the Arborium website and codebase are primarily AI; although I doubt either Amos or I have any way of confirming or denying the percentage of code that’s AI-generated.

This, along with some recent drama I learned about, led me to pass on using Arborium. Even with the bug fixes, I would rather go back to my Lezer integration.

The recent drama

This was not happening entirely in isolation. As I am wont to do, I was bitching about being unable to make Arborium work to my friend Damien. He said that he was unsurprised by the use of AI in Arborium ever since a month ago, when Amos accused zkat (the maker of the miette error-handling Rust library) of “witch hunting” after zkat added Amos to a list called “open slopware”, which I have heard was meant to list software and developers that use AI. Perhaps in part due to Amos’ comments (I am not entirely aware of the timetable), “open slopware” got a lot of backlash, which led to zkat receiving a deluge of harassment, which resulted in her announcing her leave from the open source community the day after Amos’ accusation. At the same time, Amos ripped miette out from all of their open source projects and vibe coded a replacement.

This drama is not the most relevant to Arborium as a project, and I’m not too bothered by Amos replacing a dependency for personal reasons,⁹ but I find it to be a bit childish to make AI write a replacement for a significant dependency for all your open source projects in such a short period of time. At least treat the process of replacing this dependency with the gravitas it deserves to ensure that the end user of the library doesn’t have a degraded experience from your drama.

Back to Lezer

While returning to Lezer took no effort, I could apply the lessons I learned from integrating Arborium to improve my Lezer plugin.¹⁰ Doing this saved about 100 lines of code and greatly simplified the logic. Arborium still had some features that my own plugin lacked, like using custom HTML tags to mark up the code to reduce file sizes, but it would be trivial to implement them myself.

I have come away from this with a newfound appreciation for Lezer. While it has some issues, it also has real documentation, a similar level of power to tree-sitter (with minimal concessions made for the limitations of the web), and is easy to work with. The most complicated code that I needed to make Lezer work is still simple enough to be embedded in a blog post. (It was, in fact, embedded in Joel Gustafson's Syntax Highlighting on the Web.) The only trade-off I’m making is an increased difficulty in using off-the-shelf coding themes, but my blog uses a custom-built coding theme, so that’s not a problem for now. It was also recommended by the Xi-Editor guy, if that’s worth anything.

Out of an abundance of caution, I have reached out to the developer of Lezer, Marijn Haverbeke, on whether they used AI in their code. Here is their response:

No, I don’t use language models and I’ve been updating my contribution guides to forbid slop code in pull requests.

But does Arborium work?

Even though I have moved away from Arborium due to my own personal distaste of AI-“enhanced” development, this is not something a developer with weaker moral convictions would care about. Amos fixed the bugs! They even implemented manually importing programming language grammars, something that I off-handedly mentioned that I would appreciate. If you didn’t care about any of the other drama, this sounds unambiguously good. I thought it would be prudent to see for myself before publishing this blog post, so I updated all my code to match what was described in the PR:

// Deno has its own method for reading files, but it’s what the PR said to use.
import { readFile } from "node:fs/promises";
import * as htmlGrammar from "https://cdn.jsdelivr.net/npm/@arborium/html@2.13.0";
import * as cssGrammar from "https://cdn.jsdelivr.net/npm/@arborium/css@2.13.0";

const programmingLanguages: Record<string, unknown> = {
	html: htmlGrammar,
	css: cssGrammar,
};

// The PR said to import the wasm from the node modules folder, which Deno lacks.
// I had to download it myself from jsdelivr.
const wasmFiles: Record<string, unknown> = {
	html: await readFile("./helpers/arborium/html_grammar_bg.wasm"),
	css: await readFile("./helpers/arborium/css_grammar_bg.wasm"),
};

codeElements.forEach(async (element: Element) => {
	const programmingLanguage = language(element);

	const languageGrammar = await registerGrammar(
	  programmingLanguages[programmingLanguage],
	  wasmFiles[programmingLanguage] as any, // this was the only way I could make it work
	);
	const highlightedText = await languageGrammar.highlight(element.textContent);
	element.innerHTML = highlightedText;
});

This is so much more complicated and hacky than my updated Lezer plugin. Even if it works, I would not want to use this and it would feel too brittle to not feel like my blog would break at any moment, but I’m trying to roughly match what the pull request says to do. I build, and...

I guess I shouldn’t be surprised.

To give a way to automate adding a section to a personal website which links to the websites of your friends and allies, and which does so via the medium of 83x11 badges. Ideally you'd have a list of websites in whatever tooling you use to generate your site, and when you push the button it goes off and finds the buttons and downloads them, and then it puts them in a nice little display, each pointing to the canonical url for that site with the proper alt text.

good news! the well-known/button spec already exists, just in a different form. you can find it here: https://codeberg.org/LunarEclipse/well-known-button. i know about this because beeps uses it, and because it's currently (partially) implemented on the 3.0-ish¹ branch of my site.

there is a practical difference between v's spec and the well known spec, which is the well known spec uses json while v's spec uses meta tags. i have no idea if that makes a real difference though, lol.

Silly website maintenance stuff

Unfortunately, I spend a lot of time on my website talking about making and maintaining the website. Trust me when I say that I make it look much more difficult to run a website compared to how it actually is. Even still, the website has shown a lot of positive growth this year! Not in the numbers sense, mind you. I don't track analytics. But in the past year, I've shipped Evie On-Line 2.0, which was the huge redesign I did a while back, I've added comments, and I've tried to streamline posting. This has somewhat worked out, and I've had spurts of more semi-active posting on here. Generally, I think that the biggest thing holding me back from shitposting on my own site is this faux air of prestige. I feel like I can't flood people's (RSS) feeds with my own garbage because it drowns out the quality posts and because it'll bother people. Outside of that, the site lacks real shares, and I'd like to better unify the design system and write more about the actual systems behind the site. This site has a bunch of interesting tech powering it! ...Not that it really warrants it.

Outside of that, There's the annoying conversation about hosting and morals. The code for Evie On-Line is currently hosted on Github, and the site itself, image hosting, and comments are all run through Cloudflare. Cloudflare is not great morals-wise, and I would like to self-host more of it and move away from Cloudflare whenever possible (not even mentioning that I've been paying for a cheap VPS for about 9 months now despite not even using it). I don't know if I'll ever truly be free, but there's a lot of improvements I can make right now without too much stress.

I also still have a massive todo list of minor improvements to make to the site that I've been (very) slowly working my way through. Thanks to Sage and Beeps for their feedback! I'm sorry I haven't made as much progress in that list as I would've liked. I started with the hardest task first...

To turn all this into a list of deliverables for 2026, it would be:

• Refine design system (smaller stuff like touch ups and bigger stuff like shares and improved comments)
• Refine site architecture
• Self-host site
• Move site codebase away from Github
• Add shares
• Add likes
• Improve comments (replies)
• Figure out a way of surfacing quality posts while still letting me shitpost relentlessly on main. I've seen some people separate "posting" from "articles" and I think that's for cowards.
• Continue to reupload cohost archive
• Do the list of feedback

Posting

That's right it's website stuff round 2. Last year I didn't post enough. This hasn't been due to an increase in touching grass. I've been on the socials more than ever and I should be on them less! The problem is that I haven't been writing enough. I haven't done enough shitposting OR effortposting. I have so many thoughts about dumb shit and I've let them all stay bottled up. This year I hope to change that. I plan to talk more about dumb manga I'm reading, whatever games I'm playing, any miscellaneous thoughts I'm having, etc. I think a big motivator for this for me is the fact that I have no record of what I have enjoyed this year that I could use for a year end list. There is no "Of the Year 2025" for me, and I want to have one for 2026.

Life

2025 was a year of big changes for me. I got my name legally changed, moved in with my girlfriend, got bottom surgery, and many other things. This year I'm hoping to mellow out a bit and try to build up some savings. I want 2026 to be the Year of Maintenance. Nothing new, just polishing up my life and trying to do what I can to make my life the best version of what it can be with the position I am in.

he looks exactly like if you put matpat from 2013 through a snapchat age filter

what the fuck is reverse chastity? what did they mean by this??? (i know what they meant they meant the guy gets laid a lot)

While this can be an interesting breaker that requires support to function, I’m moreso eyeing it as a way to enable ice destruction. Barriers are the most high value targets for Arruaceiras Crew, and Corsair can destroy a Pharos using 3 stealth credits and the tag + 2 credits ability on Crew. Logjam can be even cheaper. All of this depends on the set printing enough stealth support to make that package worth it, since I don’t think just Trickster Taka will be good enough to make you run this breaker. I’m not worrying about this. Stealth as a keyword is a tightly coupled bundle of setups and payoffs. Like similar synergy packages, you choose a suboptimal way of building your deck and building your rig for the payoff of getting to use good stealth cards. Obviously, Corsair is total garbage if Vantage Point stealth decks will only get to use the Ashes stealth cards, and I think the review would be uninteresting if I said, “I don’t see any stealth support from the first card spoiled for the set. 0 stars.”

On the other hand… Imagine the future where Corsair gets no support and Shaper tries to treat it like a classic Shaper HyperMantle-esque 1 breaker suite with importing Taka and using Chromatophores and Pelangi to try to make everything a Barrier. It’d be so funny.