biteshelter

I am very bad at sticking to one topic per post. At the same time, I expect myself to have one clear topic per post. My drafts are many and my posts are few. Posting on Mastodon usually gets around this because my thoughts on a topic don’t have to be complete for me to feel like posting them, but (I rarely post there either and) my usual instance has been down. So, I’m back here.

Here’s what I’ve been working on.

Since I refuse to use LLM services personally, getting a job where these things are everywhere has been a whirlwind for me. There are so many little things that surprised me about how people here use and talk about LLMs. I learned how to download and host one locally so I could poke it around for a bit and figure out what the hell people are talking about. Unfortunately for me, setting up these things has some of the same draw as customizing a website or modding a video game. I ended up getting pulled into that for a bit. You can spend so much time comparing models, tweaking parameters, adjusting prompts… After the novelty wore off, I got tired of it pretty quickly and dropped it. Despite my temporary interest, my final opinion on these things is even more negative than before.

Rant

The gap between huge models and ones that can be run locally isn’t nearly as big as I thought it was. The majority of what actually makes an LLM produce more passable output is just… making it do less. You write code and then make the LLM summarize the output of the program, or put some variables in a list, or whatever. You write something and then have it repeat it back at you because the “most effective usage” of LLMs is just to give it what you want it to output in the first place. It either does fuck all or it’s an expensive accessory to you doing it yourself. I knew it was annoyingly bad, but something about how little difference it makes to take more data and train it more irritates me for some reason. It’s just so much nothing.

The good thing that came out of all that was the urge to double down on actually doing creative things. My Fandom Trumps Hate progress has been slow but I’m now at the last stage before my art is finished and ready to send to my bidder. The zine that I’ve horrifically over-scoped is also moving slowly, but it is moving.

I also made the self-gratifying decision of learning game development with a game engine specifically for Rust (Bevy). It is said to update with breaking changes about every three months. The concepts aren’t all new to me because I’ve worked with GLFW in C++ for game development, but I’ve never done any visuals in Rust and I’ve never even heard of ECS before. Maybe it’s a horrible choice. Maybe it’ll push me to do things with reasonable scopes and time frames, since they’re all learning projects for now. I just spent a couple hours the other night developing a system where you can click on a dialogue box to read the next bit of dialogue. Most game engines probably give you that out of the box. Still, I’m proud that I figured it out myself.

In the field of note-taking, I’ve been thinking about how to better incorporate fiction into my notes. I have no real place to save character notes or notes on my personal taste. My system was designed for studying. Stories appear intact in the input and output, but the process through mangles them and leaves them empty. They can’t be converted to a series of claims and principles and still emerge recognizable as a story. My previous struggle with this was my refusal to reserve a section of my notes specifically for dealing with fiction. The Zettelkasten system is not designed for sectioning. The idea is that not separating notes by topic makes it easier to draw conclusions across multiple topics. However, I’ve since accepted that (1) a single folder full of everything is horrible to look at and (2) some things simply have their own rhythm that can be better represented with a specialized structure.

So. I now have a new folder for fiction. I’m calling it “world-building” for now, and tagging its contents as various “story elements.” I expect either for the terminology to change or for me to forget it was ever a choice. So far, working with it has already been noticeably different from the usual ZK process. The ZK is about breaking down media and experience into evidence and theories, then collecting notes with recurring themes into a conclusion. This has been more like putting down a large outline, sectioning it into segments, and filling it in piece by piece. Top-down instead of bottom-up. The whole is the primary unit, instead of whole media only being the input and output with a sea of small notes in the middle. We’ll see how it goes for creating new things instead of just migrating old story notes (from memory, because I seem to have lost much of them).

Current Mood: tired
Current Music: "Music" - underscores

My major project for the last couple of months has been a zine about some of my recent personal experiences. Lately, I’ve been feeling stuck trying to continue it.

I’ve previously mentioned on this blog that I struggle with things like identity and keeping in touch with people. While working on these things, I started putting together notes on the ideas that have been useful to me. The result was something with enough overlap and recurring points that I felt able to connect it into a model that roughly covers everything I’ve found effective.

Finding ways to put the model into practice has been helpful. However, since many of the concepts are related to plurality, neurodivergence, and queerness, I found myself in awkward positions trying to explain my experiences to people.¹ I figured it may be a good idea to compile these ideas into a single reference text. So, I started a zine.

Writing the zine has been difficult. The further I get into it, the more doubtful I am that it’ll be of any use. I started with some confidence because I knew that many of the guides I had been using before felt lacking to me. There were things I knew I wanted put into writing. After actually planning this thing out, I now feel overwhelmed by how much I don’t know, including whether putting this in a format for distribution is worth the effort. Even though there are a million self-help guides out there, and I’ve been helped by writings similar to this one, I feel like this project is a childish thing to be working on.

I want to continue this project despite my doubt. I know that I’m still proud of my first zine even though I consider it highly flawed. It seems likely that this zine will end up similarly: Even if the process is difficult, I’ll learn from it and be proud that I made something. I’m less willing to bet on other people being interested in it, but there’s a chance for that, too. I hope trying to focus on this being helpful for myself before worrying about readability or distribution will keep me motivated. I’d still like to get a sense of who I should consider sharing the zine with when the time comes, but the main priority is making sure I actually write it.

1: I help with intergroup dialogue facilitation sometimes.

As I mentioned in my last post about writing, the Zettelkasten method has been both a blessing and a curse. It's a great tool for putting together something to talk about, but it's also a never-ending writing project in itself. My approach to it has been to limit its growth. Realistically, that was never going to work. A useful Zettelkasten is a growing Zettelkasten. So, here's what I've recently learned about maintaining it.

( Letting the ZK do what it does )

The point of all of this was to have somewhere to note that the process I described previously only works for things I am currently learning. The lack of knowledge limits my main points to something manageable for a single piece of writing. I needed a card about my writing process where I could link together the conclusions from the blog post with that new information. There turned out to already be a "writing process" card in my Zettelkasten, which described a similar but outdated process that could be a useful reference in the future.

While creating the new card, I realized my current writing process mirrors my graph updates. I guess that's another way that a Zettelkasten is a writing project. Now I'm wondering if this less-straightforward method of updating a Zettelkasten, where something is created when cards are allowed to conflict with one another, may be helpful for less-straightforward types of writing.

Since it's Small Web September (and talk of Writing Month has been going around after whatever NaNoWriMo did this time), I've been wanting to start writing more often. Both my Dreamwidth journal and my new Neocities site have been a little bit neglected. I always end up blanking when I try to write something.

Problem-Solving

After strewing some writing ideas around and immediately losing track of them, I decided that I needed to find a dedicated spot to keep unused ideas before coming up with a way to flesh them out. I didn't want to make something completely new because I'd forget to maintain it. I figured that the most memorable place I could put ideas in is Logseq, which I almost always have open for notes and reminders.

Logseq is a double-edged sword when it comes to keeping track of things. It's built on the idea of "querying" notes to find things again no matter where they are, but that's because it's meant to be used for association-based (as opposed to hierarchy- or chronology-based) organization systems. Zettelkastens, which my note-taking system is based on, are meant for writing, but I had no idea how that worked in practice. This page was the most helpful guide I found. It's focused on book-writing, since that's what Zettelkastens are often for, so I tried reducing the scale of the recommended method.

I accomplished that by making a modified version of my "card" template instead of making a list of other cards. Each of my cards has space for linking to related cards, taking notes, and linking to stuff I referenced (outside sources or my own notes). I figured that I could split up the "notes" section into a premise, an outline, and the actual writing. This format seemed the easiest for me to use. Even though I'm not great at making cards consistently, they're simple enough to make and remain useful for some time. The fact that I hadn't lost them all was also a good sign. If I had an idea, I could make a card for it and come back to it later.

Possible New Problems

There are a couple potential drawbacks I can see from managing writing this way. The immediate issue is that Logseq is pointedly not for longform writing. It's bullet point only. That means using multiple programs for anything more than a couple of paragraphs, which is another step to take before completing anything. It also means deciding whether to save work in multiple files or to keep everything inside Logseq. I don't want to lose where my draft is, but I don't want Logseq to misbehave when another program edits its files.

Secondly, a Zettelkasten can be kind of a pain to use regularly. It's meant to sprawl outwards endlessly. I curb this by writing everything in journal entries and only copying significant things into the Zettelkasten proper. This writing process would require directly expanding the Zettelkasten every time I want to make some piece of writing. I don't know how large the ripple effects of that will be on my other cards. The Zettelkasten system is meant to be accepting of chaos¹ so you can offload overwhelm, but it sometimes still feels overwhelming in itself.

(The good news: By this point, I realized I had enough notes to write a blog post, and all I had to do was explain what my bullet point notes meant. And from that...)

Conclusion

The key takeaway here is that I really, really, cannot focus for shit. Journaling works for me because I either write a whole thought as soon as I have it, or I write a short summary of the whole day in one go. I lose interest when writing any more than that.²

Logseq's ability to make reference copies of bullet points was the missing link here because I could take important bits of my learning process out of my journal and directly into an outline. Fleshing out each point is just a matter of taking notes on that outline the way I take notes on anything I learn from. Writing the actual text takes much less time and focus when it's basically already done before you sit down to do it. Even though this post took a few days to write, that wasn't a death knell for the draft because of how simple it was to finish.

Well, now to link that knowledge into the rest of my Logseq notes so I don't forget it.

1: This might be what people call "antifragility." Apparently I'm not the only person to make that connection.

2: Revising, though, I could do for hours.

"Write a two-paragraph blog post for once," I told myself. "It'll be easier to get done!"

Current Music: "Suffocate City" - The Funeral Portrait

Profile

biteshelter

Expand Cut Tags

No cut tags

Syndicate

Page Summary

Check-In and Creative Progress
Doubt
Updating a Zettelkasten
Writing About Writing (and Task Management)

Style Credit

Base style: Brittle by sforzinda
Theme: Blue Days by krja

Page generated Jul. 10th, 2026 05:22 am

July 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 2026

Things That Chomp

Entries tagged with topic: writing

Check-In and Creative Progress

Doubt

Updating a Zettelkasten

Writing About Writing (and Task Management)