> There is a spooky power in having things *by heart*; knitting knowledge well into our minds changes us. — @kissane.myatproto.social, www.wrecka.ge/landslide-a-...

This widespread ”AI” conflation has set back the push for sound digital policy by over a decade. In our book about the shift to a digitized society, @ntusikov.bsky.social and I deliberately chose to avoid the term as much as possible precisely because it sows such confusion.

I wish I had the words to describe just how valuable @kissane.myatproto.social‘s latest essay feels to me: www.wrecka.ge/landslide-a-...

An easy way to explain this: Which is easier to do, bake a cake from scratch, or fix a cake that someone baked after they accidentally used salt instead of sugar?

There are of course other reasons not to use generative LLMs for drafting. But leaving aside the training sets for a moment: They're also not good at it. And they're bad at it in ways that will make us bad at it.

My experiments produced two types of results: clearly garbage, and reasonable-at-first-glance. The second is what will wreck you if you are really trying to use it as a drafting tool.

A while back after hearing a bunch of "aren't you worried AI will replace you!?” I tried a couple of experiments—with text I was never planning to use, and haven't—to understand their capabilities. The specific WAYS they're bad at writing fiction make them worse than useless as a drafting tool.

I'd argue it's worse than a blank page, in the same way that garment pieces cut slightly off the grain are worse than not having cut yet. When a thing is wrong, it will suck your energy into making it Less Wrong, which is often not the same as making it right.

it's not that I hate tech. It's that I understand the tech enough to tell when I'm being lied to

Another reason to revisit old work is that digital photography you isn’t stuck with the choices you made. Back in 2018 I was largely a black-and-white photographer, but I’ve been looking over those photos and reworking them in colour. Sometimes it works. Sometimes it doesn’t #photos #photography

I do writing and photography retrospectives to find my “voice”. Bodies of creative work have a set of consistent choices that together define an author’s voice. These four photos, taken over 25 years, have a similar voice despite being taken on three different kinds of cameras #photos #photography

Safari has started to disappear newly open tabs on me. I often open a page in NetNewsWire, press the button to open it in Safari and then switch back to NetNewsWire to continue reading. But of late, that tab is quite often gone when I switch back to Safari.

“Landslide; a ghost story” https://www.wrecka.ge/landslide-a-ghost-story/ > Our understanding is only lightly present in our own brainmatter, and exists mostly out there in our loose and jumbled substrate Great essay, one that I'm going to have to unpack over the next couple for weeks

Every “safest country in the world” survey: ‘Why is Iceland safe with the storms, earthquakes, and volcanos?’ ‘We’re an island in the subarctic with few people and no resources anybody wants, so we don’t need an army to defend the sheep and only a couple of boats to safeguard the fish’ ‘Ah’

oh my god this actually reminded me that someone *said this to me on twitter* during this. like literally, "can you even violate copyright on romance, they're basically all the same" anyway I wish I had not unlocked that memory and also that I didn't have so many similar ones to unlock lolcry

@gaudipern.bsky.social tells is like it is: LLM-based so-called AI is unpaid labourers. If you prompt an ecocide-plagiarism-psychosis radioactive waste machine, popularly known as genAI, you're exploiting unpaid labourers.

The utility of genAI—whether it can actually perform the task at hand—is irrelevant and OSS ultimately a trap to bother talking about it. There is literally no amount of usefulness it can add that could make it worthwhile. None. Willing users are either ignorant or willfully choosing massive harm.

There is no such thing as “harm reduction” that can apply to this technology. Its social, political, and economic repercussions will do irreparable damage to actual humans. You know, the ones we should care about? But whatever I’m mostly just saying this so you know why I think less of you.

Just a thing about this: a common statement in writing is that you can fix crap, but you can't fix a blank page. I have heard AI apologists claim that this is why it's fine to generate a first draft. Now, to be clear, fuck this. But also: an AI draft is a fucking blank page.

This seems a good litmus test for LLMs you will not be able to replace your devices - be they of the desktop, laptop or handheld variety - but the devices you do happen to have at the moment have the full range of chatbot functionality is it worth it?

A great review. Geoff and crew have done so much to keep an absolute pillar of our industry alive and flourishing. Hate the see how damaging Google “AI summaries” have been, but love to see how much is being invested in the CSS Almanac. It’s one of my favourite resources on the internet!

Cunk spittin Christmas cheer

The news about Harlequin translators getting fired en masse? You should care about that. They're doing it there because they think nobody gives a shit about romance and so it's fine to experiment with it first.

The news that AI was being trained on stolen works? It was broken first about the BooksCorpus, on May 11 of 2021, and it was discussed in major news media thereafter. Romance authors talked about it. A lot. The first archive they stole was largely romance. arxiv.org/abs/2105.05241

“Write to the market.” You’ll never catch the market. By the time you’ve written it, the market is gone. Write what you’re on fire to write, and sometimes the market shows up.

Saying it again, anthropomorphized "chatbot" user interfaces to LLMs should be illegal. There's zero reason these interfaces have to be like this—other than companies making people addicted to them.

Whenever someone suggests lore matters at all, when the point of storytelling is dramatics, I get annoyed

y'know why The Raid rocks so much? because we only know a few important things about Rama (he's a rookie with a wife, kid, and estranged brother) and then puts him into the most insane, unimaginable situation possible what matters is the dramatics, NOT A SINGLE CHARACTER'S BACKSTORY

Firefox browser falls to AI. What do we do now? where ‘absolutely not’ means ‘maybe later’ www.youtube.com/watch?v=c40i... - video pivottoai.libsyn.com/20251223-fir... - podcast time: 9 min 39 sec pivot-to-ai.com/2025/12/23/f... - blog post

Some of you have forgotten that only three years ago you were perfectly capable of writing a text, writing an email, telling a bedtime story to a child, and it should worry you that powerful companies have convinced us we can’t do things we’ve been doing since the dawn of time.

Tonight Kertasníkir (Candle Stealer), the last of Iceland's 13 Yule Lads, comes down from the mountains. He steals candles (that used to be made from animal fat) and eats them. Children leave their shoes out in the window, he gives the good ones a small gift, the bad ones get a potato.

How is this legal? And if it is legal, what exactly do we take a torch to it legislatively? www.washingtonpost.com/lifestyle/20...

You know, we're always told we view the past with rose-tinted glasses & the olden days were never really that good, but I remember when journalists used to piss & moan about having to do human interest stories instead of hard news because there wasn't enough news to fill half an hour. I miss that.

This is the gist of what I’ve heard from people who do use AI / LLMs at work (outside actual data handling). Its best workplace use is to automate stuff that nobody should be asked to do in the first place. Automated time-saving generation of the things no human ever looks at.

love that software developers are advertising AI functionality in every single app now and it's up to you to try to figure out if they mean "this has spellcheck" or "everything you create in this app will sync in real-time to sam altman's personal laptop" or something in between

by far the funniest use case for AI is “I use it for powerpoints because I hate making them and nobody looks at them” because it kinda sounds like the real workflow savings would be to stop making powerpoints

Very good paragraph here from @cstross.bsky.social on LLMs… www.antipope.org/charlie/blog...

This is a wonderful summation of the beauty and value of human made art ❤️

One of the things I’m trying to figure out with my impromptu photography retrospective over the past few days is a better understanding of how I can deal with colour. I’ve been trying to figure out when colour is necessary, when it adds, and when it detracts, how best to use it. 1/6

We begin, appropriately enough, at the end. Where everything is revealed and some things are put to rest. #12daysofYule #thedevilssound #witheredhill

PREACH IT TECH JESUS

bsky.app/profile/pook... The einstellung effect has been known for generations: you have a very hard time noticing you're not actually solving the problem you actually have, but a different problem. Generative AI is the einstellung effect reified.

Learning how to pay attention to small details, how to deliberately strain yourself to notice tiny details that are off, is a skill programmers learn the painful way. And, using a chatbot deskills you in this very extremely important skill of being able to pay attention to small details.

hours of tracing an error only to bolt awake in the middle of the night - your subconscious having clocked away to find a solution. denying yourself those moments of joy and discovery leads only to copy-paste limbo with zero innovation

as a black tech anarchist who hates closed-sourced applications and the concept of BIG TECH, the last 5 years of my existence have been so validating for me

Somewhere out there right now is a programmer vibe-coding an FDA-regulated medical device, and somewhere else is the unemployed person whose job was to inspect that code and hold it to standards and who was fired by a 19 year old.

Also, this information isn't in any training module... cuz I haven't published it yet! And there's no other way to learn this except in the archives where I read the panicked letters between Lougee and the IT department! There is *so much* history AI does not know! 🗃️

It's good that more attention is being paid to the plague of AI-drive scams that is driving writers batty (but please, don't mention Writer Beware, which has been reporting on them since August) www.thebookseller.com/news/authors...

obligatory plug of this piece @kara-williams.bsky.social and I wrote a few months ago explaining the stakes of policymaking with vague/improper ideas of what AI means consumerfed.org/specific-ter...

I believe an entry-level software developer's learning journey should include diverse outside influences because it would be very bad luck for about 90% of them if it didn't. "We want them to learn how we do things here." And then, hopefully, they get to learn better ways than that.