skip to main content

mini-posts

about mini-posts.

these are miniature blog posts that don’t clutter up your rss feed. perhaps they are seedlings in my digital garden; perhaps this is just a way for me to pretend i don’t secretly pine to post on social media once more.

there is no way for you to know if there are new mini-posts yet. mostly i built this as an in-between my mental notes, my paper notes, and a full-blown blog post.

2025-07-13

a post about LLM

a corollary to my previous mini-post here: i think that the way we are talking about what people try to do with LLM is generally wrong, with most scholarship around the matter approaching it from the angle of “automation”. i can speak to coding here: some of LLM-work can look like that (boilerplate generation is a commonly cited one, but note that this has existed in many forms prior to LLM) but a lot of it seems to be about reducing cognitive load, which i don’t think is really automation per se.

another analogy i’ve heard is that LLM gives you leverage (akin to a forklift, the invention of which, combined with palletization and containerization, vastly standardized logistics to make it an infrastructure). is LLM akin to a table-saw, which can accelerate work but can also be extremely dangerous if you don’t know how to use it? again i don’t think this is quite correct for reasons i admittedly can’t quite articulate (hence me working it out here on the mini-notes.)

rather, i think that LLM (and generative ai in general) are attempts to approach work from the other direction. it’s not so much “create an essay from these bullet points” as much as it is “i want an essay that summarizes down to these points i gave you.” it explains why the buzziest form of “vibecoding” involves services like lovable, replit, or vercel—all frontend-focused services which, from my experience, create extremely unpolished frontends. eevee calls LLM work artifacts whatever; i prefer to call it bullshit.

credit where credit is due, i suppose. but you might say: ms. inwiring dot com, what about SQL queries or whatever? to which i reply: sure, natural-language querying is a tool (which existed before LLM, mind you) that has probably benefited from LLM (though i love SQL) but people are also using LLM to do things other than natural-langauge querying. they see that problem domain and attempt to parlay it into other problem domains. it’s also notable that in the SQL case, the final artifact is not usually the query, it’s usually not even the data, but some sort of report. of course, you can use the LLM to make that report now, too.

we live in a work culture that has only ever valued the artifacts of output, while forgetting that artifacts exist as proof of work. this is why i’ve been so pessimistic about LLM:

  1. david graeber points out that a lot of “unskilled” work is in fact caring work—think of a housekeeper, for example, who must learn a lot about your living habits in order to be successful at their job. but i think a lot of knowledge work is also caring work. think of the intern who produces slide decks for some investment bank, a job which is apparently “automated” now through slide deck generation tools. i am not saying that producing slide decks was worthwhile work (in fact i think it’s a textbook “bullshit job”) but i am also saying that slide deck intern jobs are there because companies need to train junior employees, junior employees need to learn how to do their job, but training is so devalued that bullshit slide deck internships were the only way to sneak that in. in producing those slide decks, the intern gets an idea of how to do their job, if they are to progress up the corporate ladder. that’s caring work in a sense, and now we have a technology that promises you no longer need to care at all.

  2. and yes, let me apply the jevons paradox here: let’s assume that it’s true that LLM reduces cognitive load to, say, write a whole fullstack app yourself, as many have claimed. the app is the artifact, but remember, the artifact is not the work. so if the software industry reorganizes around these tools, then it is certain that engineers will be tasked with even more scope and even more responsibility. the consequences of this are surely that work that was once necessarily collaborative becomes siloed and solitary. if you look at the trees, maybe you see automation. but if you fly up and look at the forest, i see a grim future for knowledge work: needlessly duplicative, foisted onto an ever-shrinking cohort of workers, who tackle an ever-increasing scope of tasks without ever truly understanding what they’re building, just slapping fresh coats of paint onto crumbling infrastructure forever.

a few more thoughts:

  • i think the LLM-centrists who say that this is “just a tool” are wrong on multiple counts: for one, LLMs aren’t really tools, but people use them to make their own tools. “agentic” LLM is proof of this, it’s an attempt to scaffold entire workflows.
  • this is why i also am extremely skeptical of protocols such as MCP: to me, they are mostly just rebadged REST APIs. but they come with a lot of security downsides, so why wouldn’t you just write the API client yourself, or, you know, use the LLM to boilerplate one for you? MCP is evidence to me that the thing we are chasing isn’t actually automation.

addendum

if you’re interested in infrastructure i love how infrastruture works by deb chachra.

also, these “mini” posts are getting quite long, huh? time for me to collect these and turn it into an article.


2025-07-12

a particular bugbear of mine has been the way some describe neo-Luddism as “reactionary”. this is one of these annoying semantic shifts that have been caused by context collapse, imho: “reactionary” politics is against labor, against socialism, against equality, and for existing power structures. it doesn’t mean to be reflexively against something.

as a luddite i’m concerned about further increasing my dependency on cloud computing, which due to primitive accumulation all but guarantees that ownership of compute—not “hard work”, not ingenuity—will predict the power structures of computing (and if computing is to be further integrated into our lives and thought processes, ourselves.) i do not see why one should fully embrace technology pushed by billionaires who clearly want to entrench their positions of power!

that is not to say that technology cannot be involved in social change. i am trans! i take hormones, a technology. hormones have undoubtedly changed my life, and are a miracle of scientific innovation! but now hormones are the site of contemporary gender politics, and my access to hormones is constantly threatened by politicians (backed by these very same billionaires, btw). and in this we can see attempts to skirt around mainstream power structures through technology (diy hrt providers, which often accept cryptocurrency—a technology that is generally not anonymous). if you’re on this boat you should absolutely be at least somewhat resistant toward artificial intelligence, which we are already seeing being used to analyze massive data sets at scale, by both the IDF and by ICE. you cannot outcompete in a rigged game, because the billionaires will bend the rules to acquire even more resources.

no, at the very least you need alternate social power structures. (which if you wanted to take the piss out on me you could say is a technology.) but that’s what i fear the most about chatbots (and what i fear is already happening among people who interact heavily with them): if your emotional and social skills atrophy, you will become less competent socially, and you will become more emotionally pliable and dependent on a computer that doesn’t care about you. for one thing: social isolation and hyper-individualism, a tension which i already have felt in my work. i have already seen the way social media and the associated advertising tech warp how we see the world in bad ways and i am honestly not ready to go do the second round of this.


it’s not very often that you see the social construction of technology happen so distinctly. a little over ten years ago, one big piece of tech news was that users taught a chatbot to be racist on a social media platform, which resulted in that chatbot’s shutdown and the relegation of chatbot technology into mere curiosity. a few days ago, the owner of that same social media platform has taught a chatbot to be racist and antisemitic (in the name of “truth”) and its users have directed the chatbot to sexually harass women. and now? nothing: the chatbot has become an inescapable aspect of online life, so it seems like one must just deal with it.


2025-07-11

i was gone for a bit and touched grass, and apparently in the ensuing moments a lot of stuff happened.

one of those things was a discourse on bluesky about free buses, something i have advocated for in the past, perhaps stridently. i think it’s good policy, period.

i can only speak for nyc transit, where the bus is a secondary form of transit to the train, and where most people outside of staten island get around by transit. from what i recall, most of the scholarship around “free transit” look at jurisdictions like luxembourg where fares were dropped in the hopes of promoting transit modeshare among a population that mostly drove. the goal of the free bus program are clearly not to promote transit use per se, but rather to reduce the costs of bus commuting in communities not served by the train. it’s not for nothing to point out that $6 (the rounded-up cost of two bus trips) is a lot of money actually. it can buy you lunch!

free fares also means that the “crime” of “fare evasion” is categorically eliminated, which means less conflicts with transit employees and less time wasted on something that has no effect on public safety.

fare free buses means no time spent collecting fares, which opens up opportunities for multi-door boarding, which can speed up buses.

and finally, i think that the richest city in the world can afford to allocate more funds toward operating a public good like transit! i think the moral argument for free fares is a great one actually.

hmm. perhaps i should go back to touching grass, actually.


2025-07-06

there’s been a lot of malignment of the em-dash recently. (some of it from me!) so i’ve tried to use it more consciously. as a form of punctuation, it really helps paragraph text flow—especially as we now read on variable-width screens. but overusing em-dashes (or bullet points, or even clichés such as “X is not just A, but B”) all actually have something in common: they are cheap ways to break up and altering the pace of the writing.

who am i to rally against cheapness per se? if it’s cheap and does the job, why care? and yet, no one decides consciously to turn good things bad. rather, they just cut one small corner one at a time, and they get used to that so they cut more, and more, and more. do not confuse economy with cheapness.


i feel like my social anxiety has been supplanted by an overall Weltschmerz which, um, seems bad? perhaps i am acclimating to my lack of control (as an individual) in this world. no one really does exert full control over their environment: otherwise, the tyrants who run the world would be able to conceal their insecurity and their bereftitde of interiority.


2025-07-01

i’m not sure who reads my blog (possibly no one), but i’ve decided that if you do, i want you to feel like it’ll be worth it to dedicate your attention to a blog post. this is not really a new idea at all—a whole variety of blogs did this—but in each new post going on forward will be a plain summary that i think accurately describes the contents and mood of the post following it.

this isn’t a “tl;dr” (which is more like putting the conclusion of a piece of writing at the front) nor is it a “teaser” (i want the summmaries to be the opposite of engagement bait) but it’s also not meant to be an abstract. think of my philospohy behind summaries as signposts toward a destination, should you proceed to read the post. and needless to say, i write these by hand.


2025-06-30

there’s an interesting phenomenon of inversion that happens with language, at least the english language of the united states. “awful”, formerly a synonym of “awesome”, became its antonym; “lifting yourself up by your bootstraps”, an impossible act, became a common utterance in the united states and the driving philosophy of american conservatism; nonperformative speech has most recently become re-labeled as “performative”—in other words, the term “performative” became its precise opposite.

there are probably so many other examples of this phenomenon, but the “nonperformative” to “performative” shift has been particularly pernicious because it’s distorted the notion of performativity (i will elaborate on nonperformativity in a later post, i hope.)


my collection of anecdotes of people using a technology are trustworthy because they’re from competent experts at the technology, and are encouraging signals of emergent behavior, in fact. but, you see: your collection of anecdotes of people developing severely adverse effects from using a technology are actually invalid because those people using it are incomptent at using the technology and perhaps predisposed to incompetence. plus it’s behavior one can simply patch away.


2025-06-17

thinking of moving this away from neocities tbh. i feel like i am cluttering up a global activity feed somewhere, plus the toolchain to deploy this is both anti-neocities and baroque


i have no autocomplete and i must scream


in ceramics, tap centering (a somewhat stochastic process) is, once mastered, a great boon to keeping you in a flow state when it comes to trimming ceramics. you have to be prepared to completely reset your piece if it’s about to fly completely off the wheel.

so i learned to tap-center. it’s pretty cool. it helps me a lot, personally, because i think i have poor spatial reasoning. but of course, some of the most talented ceramicists i know don’t know how to tap center or haven’t bothered to. flow isn’t everything, and intention matters just as much, if not more.

archive

2025-06-14

why don’t they just put a clearly labeled sign on chesterton’s fence describing what it does and why you shouldn’t take it down?


2025-06-13

alan kay’s original vision is that a computer is a bicycle for your mind. it’s a way to amplify your natural intellect and go further.

i like this analogy a lot, because bicycles also happen to be very friendly for the environment. i don’t just mean the climate, i mean the built environment as well. a trip to the netherlands should prove this to you.

but bicycles still have externalities. (for example, also in the netherlands, i believe it’s common to leave two or three bikes around town, which clogs up bike parking, which is why governments are now promoting bike share and transit as well as building out some impressive bike parking facilities.) plus, sometimes i don’t want to bike. which leaves me to think: what is public transit for your mind?

a library, perhaps? a community of interested computer users? a communal computer like dynamicland?