Baldur Bjarnason

Writer, web developer and consultant based in Hveragerði, Iceland. Lapsed Interactive Media Academic. Webby Tech Stuff and webby book stuff.

Blog: https://www.baldurbjarnason.com/
AI Book: https://illusion.baldurbjarnason.com/
Pronouns: he/him
Book: https://softwarecrisis.baldurbjarnason.com/

Public Key

npub1hr7087sk4yxl25nmx9c42p0zgk5kyfyvar5xeklezs23hjuenr7qr3jxyc

NIP-05 Address

[email protected]

Profile Code

nprofile1qqst3l8nlgt2jr042fanzu24qh3yt2tzyjxw36rvm0u3g9gmewve3lqpqqxpvrvz

Publishing to

Author Public Key

npub1hr7087sk4yxl25nmx9c42p0zgk5kyfyvar5xeklezs23hjuenr7qr3jxyc

Show more details

Last Notes

2026-04-29 11:41:52 UTC

Apparently, subsidised code automation tools ("coding agents") that make it 10x easier to generate code is a big part of why GitHub has been having issues
So, it really is their own fault since GitHub and Microsoft are pretty much ground zero for offering LLM-based coding tools and promoting the bubble, it's just not breaking for the specific reasons some of us have been thinking
https://xcancel.com/kdaigle/status/2040164759836778878#m

2026-04-28 12:02:07 UTC

After spending quite a bit of the past couple of weeks chatting with friends and family, asking them what they're taking from the news, I think it's clear most of our news media is substantially downplaying the economic effects of the Hormuz straight crisis
What I can't figure out is whether it's intentional or not. I know it's unintentional in Icelandic media as they simply no longer have the resources or expertise to properly outline what's happening, but I can't figure out UK or US media

2026-04-14 13:59:33 UTC

RE: https://dair-community.social/@emilymbender/116397605266279416
It's a bit of a trope to say "this is an important talk" when somebody tackles an sombre topic, but this is in genuinely an important talk on the acceleration of dehumanisation enabled by "AI" and how we can resist it, all wrapped in a decent overview of the discourse so far, from a leading expert in the field
#note1e4q…nx43

2026-04-13 11:44:53 UTC

- reply

If you warn about a problem before it's FELT, people will ignore you even after it hits them
A warning with enough foresight to precede the catastrophe will be less credible BECAUSE it was early.
People would rather be angry than feel responsible and an early warning proven right takes that away

2026-04-08 18:00:32 UTC

- reply

(Also, they will never apologise for lying to you. You’ll still be waiting for an apology at the heat death of the universe.)

2026-04-08 17:55:55 UTC

- reply

Some of y’all are middle-aged. You should remember hearing outright falsehoods from an authority figure about a financial bubble at least once in your life before. This bubble’s bigger than the others which means a lot more people will fool themselves into thinking that this time it’ll be different

2026-04-08 17:49:45 UTC

The one iron rule of late-stage financial bubbles is that with massive amounts of money at stake, a lot of people will be lying a lot of the time and many of lying liars will be people you’d trust in normal times because the first person they fool is themselves
Be extra sceptical.

2026-04-07 13:04:16 UTC

- reply

Or, to use a different bubble as an example, no matter how strongly you believed in the web, long term, talking up dot-com stocks isn’t the tail end of that bubble would have been fundamentally irresponsible and, indeed, would have been a big red flag to others signalling that you are a major unrepentant asshole

2026-04-07 12:46:33 UTC

- reply

Anybody who, during a trillion dollar financial bubble coupled with an equally seismic political movement, routinely picks up the plausible-sounding positive "AI facts" published by people who stand to gain from either bubble or movement is doing both us and themselves a disservice
Even if you truly believe your positive opinion on "AI" is warranted, giving these ideas a platform is equivalent to hyping up sub-prime mortgages in late 2007. It's fundamentally irresponsible even if you were right

2026-04-07 12:34:30 UTC

RE: https://toot.cafe/@baldur/116300859215261350
For reasons, I'm reminded of this thread from a couple of weeks ago that I wrote in an attempt to remind people to be extra-sceptical of bubble-inflating news coming from those who benefit from a financial bubble, even if it looks plausible
#note16t3…47lm

2026-04-06 22:17:12 UTC

That feeling the UI sets a date range exclusive of the end day but you always think of date ranges as inclusive, only to realise at the last minute that you need to add a day to get the actual range you want

2026-04-06 22:00:14 UTC

- reply

Stripe has a set number of countries it supports for sellers and anybody outside of that is told to incorporate in the US.
I'm not going to incorporate in a failing state that's in a trade war with the world and an actual war with a chunk of it, just to sell a few ebooks.

2026-04-06 22:00:10 UTC

This is the reason why I jumped ship to Payhip, even though it's arguably not quite as good, as soon as Stripe bought Lemon Squeezy. Iceland's not on the list of supported countries for sellers for the service that's going to succeed Lemon Squeezy
https://www.lemonsqueezy.com/blog/2026-update

2026-04-06 12:37:30 UTC

- reply

To those who aren’t “AI”-pilled, Steve Yegge on anything related to LLM coding makes about as much sense as the Roko’s Basilisk nonsense. If you want to be convincing to outsiders you need to stop citing what are effectively “AI” catechisms

2026-04-06 12:24:14 UTC

I try to keep track of what’s going on among those who use LLMs for coding but they all keep linking Steve Yegge and I just can’t take anybody who links to Steve “gas town” Yegge seriously.
He’s the opposite of convincing.

2026-04-06 09:48:33 UTC

I did not know that Daði Freyr had moved to Hveragerði. https://www.ruv.is/frettir/menning-og-daegurmal/2026-04-05-oradi-ekki-fyrir-thvi-ad-hann-myndi-sjalfur-ferdast-sem-tonlistarmadur-471690
(Hveragerði has a bit of a history and reputation for being an artsy town)

2026-04-05 16:38:53 UTC

- reply

Roughly, yup 🙂

2026-04-05 12:40:36 UTC

On days like these I'm reminded of one of my granddad's favourite words: "helgislepja". Translated as "sanctimony"
"Slökktu á helgislepjunni í útvarpinu" > "Turn off the sanctimony in the radio"
That doesn't capture how funny the word is though: it literally means "holy phlegm" or "saintly mucus"

2026-04-05 11:21:52 UTC

- reply

Ah, yeah. "transfer of values" is a useful way of describing it.

2026-04-05 11:03:34 UTC

- reply

That's pretty likely, yeah.

2026-04-05 11:00:33 UTC

- reply

There is also not much to figure out for the rest of us. The technology is purpose-designed to remove people from the equation, much like a handgun is purpose-designed to remove people from existence. Any "figuring out" about either tech will only result in variations on their purpose
What we need to do is strip back the tech, go back to the drawing board, and figure out how to reinvent it practically from scratch to be more human. /fin

2026-04-05 10:58:39 UTC

- reply

There is no rebuilding, no constructive potential for "AI" without political reform, both in the US and in Europe. The motivation behind the tech still remains: powerful people want to take things away from society
The people who funded and drove today's "AI" will have the resources to figure out how to make the tech affordable after the financial bubble pops. They're the ones who will have the resources to figure things out about LLMs, not us.

2026-04-05 10:57:00 UTC

- reply

They were built for this goal—fewer humans, less accountability—and will always be problematic no matter who is in charge and always always be a risk as long as the "shitty people" have power
In the modern history the "shitty human beings" have always retained power and influence after bubbles pop, from Reagan onwards. The bubble builds up wealth and power, they keep it after it pops, and use their influence to get in on the ground floor on the next one.

2026-04-05 10:55:32 UTC

- reply

Generative models and automated decision-making systems are political projects. They are tools for fencing off sectors of our society for rent, for cutting back on education and healthcare for the poor, for removing accountability. They are inherently tools for removing humans from the equation. They are not neutral in their design. Their existence has a political purpose

2026-04-05 10:54:49 UTC

I've been seeing people say the harms of "AI" comes from shitty people, not the tech and that the shitty people will go away after the bubble pops. While the first part is true in the strictest sense (contrary to claims, these are not autonomous systems) it's not really true in a practical sense. The harms are designed in and the "shitty human beings" won't go away.

2026-04-04 11:37:48 UTC

Cat photos for Easter #caturday
A colour photo of a black and white cat sitting in contemplation about why our spring is covered in snow.
https://assets.toot.cafe/media_attachments/files/116/346/234/533/920/740/original/cdedf424be948c6f.jpg
https://assets.toot.cafe/media_attachments/files/116/346/234/543/191/311/original/35b1a598d03a73a5.jpg

2026-04-04 10:34:49 UTC

- reply

More about the preview
https://www.baldurbjarnason.com/2026/easter-sale-and-the-toy-factory/#also-included-a-preview-of-my-upcoming-book

2026-04-04 10:34:23 UTC

The preview to The Magic Toy Factory is the first of my ebooks where I use Typst (via pandoc) to typeset the PDF version and it’s honestly been excellent.
https://typst.app/

2026-04-04 10:08:01 UTC

“Miseducative Experiences”
https://2ndbreakfast.audreywatters.com/miseducative-experiences/
> But to frame any opposition to technology as a "moral panic" is a rhetorical sleight of hand in which one side gets to invoke "science" and "research" while dismissing the other as mere "hysteria."

2026-04-03 16:07:33 UTC

- reply

🙂 👍

2026-04-03 15:54:32 UTC

- reply

No, I haven't sorted out bookstore distribution yet (though I'm hoping to get that done this year, they mostly need some format adjustments to work well with trade distribution Print-on-Demand systems). Until then, they're only available via the Lulu bookstore
https://www.lulu.com/search?contributor=Baldur+Bjarnason

2026-04-03 15:42:10 UTC

RE: https://toot.cafe/@baldur/116330178975566449
This is still going on 😉
#note1pw8…awrf

2026-04-03 15:41:21 UTC

- reply

Same.

2026-04-03 12:08:42 UTC

This six part series (link goes to the first part), written by a former core Azure engineer, is mind-boggling. Microsoft sounds like a dysfunctional company whose software is dangerously unreliable
(I mean, more so than it has been historically)
https://isolveproblems.substack.com/p/how-microsoft-vaporized-a-trillion

2026-04-03 11:29:02 UTC

Redpolls seem to be in good health despite the weather #birds #iceland
https://assets.toot.cafe/media_attachments/files/116/340/533/464/158/029/original/bf95180f36161206.jpg
https://assets.toot.cafe/media_attachments/files/116/340/533/478/102/029/original/24125c7ce32ed14c.jpg
https://assets.toot.cafe/media_attachments/files/116/340/533/478/234/893/original/42aa521462eb5675.jpg

2026-04-02 22:28:42 UTC

So, AFICT the “reasonable” argument for LLM coding is if we persist through economic chaos and environmental crises while sacrificing the creative industries and education then models built using stolen data and license violations might safely revolutionise software dev productivity?
Public sentiment against “AI” is sour enough already. If you want to make software developers as a class as broadly disliked as tax auditors, this’d be the way to do it.

2026-04-02 16:41:44 UTC

- reply

True!

2026-04-02 16:24:47 UTC

- reply

Awesome 👍

2026-04-02 16:19:03 UTC

- reply

Ah, thanks! 🙂

2026-04-02 16:18:52 UTC

- reply

Such fun 😁

2026-04-02 16:16:05 UTC

So, turns out Angine de Poitrine (Pectoral Angine? Chesty Angine?) is pretty much exactly the sort of thing I enjoy.
https://www.youtube.com/watch?v=0Ssi-9wS1so

2026-04-02 14:06:24 UTC

- reply

I've replied in email 👍

2026-04-02 11:33:27 UTC

Starting to see more LLM-using acquaintances who've spent the past three years ignoring every problem solemnly talk about "having concerns about the technology". Not many, but more.
Kinda wondering whether it's a critical mass thing (easy to ignore individual problems but not all the problems everywhere) or whether there was a single specific thing that soured them on it, leading them to see it differently.
But also don't really care why or how, just that they're finally listening.

2026-04-02 10:44:13 UTC

- reply

Haha! Pretty close 😆

2026-04-02 10:02:21 UTC

- reply

Yup. I agree

2026-04-02 09:58:00 UTC

“Mouthwords | everything changes”
https://everythingchanges.us/blog/mouthwords/
> We need to see the advent of workslop in the context of the technological aims of the last several decades, one of which has been to obfuscate the human labor involved in everything from driving to cooking to gathering

2026-04-02 09:36:56 UTC

All the devs saying that Anthropic’s code quality is “normal” are telling on themselves and everybody they’ve worked with
(Also supports what many have been saying about software quality being a crisis that precedes LLMs, but that’s another story)

2026-04-02 09:10:28 UTC

It’s April
https://assets.toot.cafe/media_attachments/files/116/334/330/912/549/169/original/b1b621ad3a4bcfb9.jpeg

2026-04-01 18:35:49 UTC

RE: https://toot.cafe/@baldur/116330178975566449
I'm doing both a sale and an experiment.
#note1pw8…awrf

2026-04-01 17:26:11 UTC

- reply

Sure, send me an email ([email protected]) with information about what you need and I'll see what I can do. If the ecommerce system doesn't have an invoice that fits, I can always put one together a little bit more manually.

2026-04-01 17:14:28 UTC

- reply

😅

2026-04-01 16:50:16 UTC

- reply

😆

2026-04-01 16:37:02 UTC

- reply

Absolutely!

2026-04-01 16:36:50 UTC

- reply

Ah, DNS problems :-( Those are notoriously hard to solve since I'm not seeing any of the issues on my end.

2026-04-01 16:35:44 UTC

- reply

Yeah, this has been Ed Zitron's beat for a while. Seeing law firms call it out specifically as a systemic litigation risk does a good job of underscoring the seriousness of the situation as well.

2026-04-01 16:16:59 UTC

- reply

Definitely looks that way.

2026-04-01 16:12:38 UTC

- reply

Yeah, the memo covers that as well in a way that's more readable to a lay person like me than I expected.

2026-04-01 16:10:23 UTC

- reply

> The interaction between the two collateral layers creates a cascading risk. If GPU collateral loses value, the neocloud tenant cannot service its debt, which impairs the tenant’s ability to make lease payments to the data center SPV, which in turn undermines the cash flows backing the data center ABS that investors are relying on.
and
> ...if it materializes at scale, would propagate losses across every layer of the financing stack.
Oh joy.

2026-04-01 16:07:47 UTC

- reply

At the very least. We have other potential crises on the boil as well 😅

2026-04-01 16:07:10 UTC

- reply

https://www.quinnemanuel.com/the-firm/publications/client-alert-emerging-litigation-risks-in-financing-ai-data-centers-boom/
> In a role reversal, this year Moody’s has warned that AI data center investments may be riskier than they appear, as current U.S. accounting rules allow companies to conceal tens of billions in potential liabilities. Moody’s noted that reported liabilities may understate likely cash outflows
No worries there, right?

2026-04-01 16:03:29 UTC

- reply

https://www.quinnemanuel.com/the-firm/publications/client-alert-emerging-litigation-risks-in-financing-ai-data-centers-boom/
> Data center debt has increasingly been repackaged and sold to a wider pool of investors through asset-backed securities (“ABS”) and commercial mortgage-backed securities (“CMBS”).
This has never backfired, right?

2026-04-01 16:03:06 UTC

This law firm memo provides a fairly detailed overview of the risks inherent in the "AI" finance bubble and it doesn't look good
https://www.quinnemanuel.com/the-firm/publications/client-alert-emerging-litigation-risks-in-financing-ai-data-centers-boom/
> The deeply interconnected AI ecosystem means that distress at any single node—a construction delay, a tenant default, unhedged energy cost differentials, a collapse in GPU resale values—can propagate across multiple counterparties and financing layers.
This is fine, right?

2026-04-01 15:34:52 UTC

- reply

If you want to see the 80 page preview and follow the progress of the book, you can buy the preview separately here:
https://payhip.baldurbjarnason.com/b/D70UH

2026-04-01 15:34:22 UTC

I have an Easter sale going! First off, a bundle of all of my books, including the ones no longer available for sale elsewhere. It's more than 60% off the full price
https://payhip.baldurbjarnason.com/b/QHWZc
But the interesting part of this is an experiment where, included in the bundle, is a preview of the book I'm working on. Buying the bundle, or the preview separately, lets you join a newsletter where I'll be documenting my progress on the book.

2026-04-01 15:12:00 UTC

“Dissolving the social”
https://tante.cc/2026/04/01/dissolving-the-social/
> “AI” exists to disenfranchise labor. That’s what it’s for.

2026-04-01 11:37:41 UTC

- reply

Too true!

2026-04-01 08:55:23 UTC

Probably for the best
https://assets.toot.cafe/media_attachments/files/116/328/606/223/756/677/original/ab4521c3e8414e62.jpeg

2026-03-31 23:57:19 UTC

Ugh. Forgot that it’s the internet’s least funny day of the year tomorrow
Though given the state of the world, getting annoyed by the web’s yearly turn towards the unfunny is a bit like obsessing about your wet socks while you’re hanging onto a cliff’s edge by your fingernails.

2026-03-31 20:59:15 UTC

- reply

I don't know how this will pan out. I don't even know the exact shape of the final book. But I think this might be a fun experiment for both me and my readers. It'll force me to do a better job of documenting my process and progress. And it'll show you some of the rough, nitty-gritty, parts of putting together a book like this.

2026-03-31 20:57:10 UTC

"A massive Easter sale and a preview of my next book (out late 2026)"
https://www.baldurbjarnason.com/2026/easter-sale-and-the-toy-factory/
So for this Easter I'm making an experiment
Instead of trying to find time in between projects to work on my next book, I'm attempting to fund that work by offering a preview of it as a part of a Easter sale bundle
Those who buy the bundle or the preview also get to join a newsletter where I document the progress of the book over the coming months and, once it's out, a copy of the final version

2026-03-31 20:56:08 UTC

A massive Easter sale and a preview of my next book (out late 2026): https://www.baldurbjarnason.com/2026/easter-sale-and-the-toy-factory/

2026-03-31 16:50:09 UTC

- reply

That's interesting, thanks!

2026-03-31 15:36:45 UTC

- reply

The worst case scenario here is Götterdämmerung/ragnarök so let’s just hope it doesn’t come to that.

2026-03-31 15:33:41 UTC

- reply

AFAICT and very much not an expert, but some of the places to watch for the really serious warning signs: India (food), Australia (energy-intensive mining), Japan (plastics), South Korea and Taiwan (semiconductors and other industry). Shortages are likely to be a much bigger issue than the market price of oil

2026-03-31 15:12:42 UTC

Trying to read up on and comprehend the economic impact of certain current events is outright nightmare fuel. There’s a non-zero chance this will play out in a substantially worse way than the pandemic 😬

2026-03-31 14:44:53 UTC

- reply

If LLMs were only used by a small number of experienced devs working with well-engineered guardrails, we'd have less of a problem
But once they start getting more commonly used, they start to pollute the entire ecosystem and the only way forward is stiff regulation for everybody

2026-03-31 14:44:46 UTC

I keep coming back to the leaded petrol analogy for LLMs and coding
Harms that are manageable when it's only used by a small number of experts become catastrophic pollution when it's used broadly throughout society
If LLMs were only used by a small number of experienced devs working with well-engineered guardrails, we'd have less of a problem
But once they start getting more commonly used, they start to pollute the entire ecosystem and the only way forward is stiff regulation for everybody

2026-03-31 14:25:22 UTC

So, still digesting the argument of this article and haven't quite made up my mind about it.
But the fact that various forms of self-aware reasoning towards problem-solving exist across the animal kingdom that doesn't seem to be a direct function of neural network size would seem to indicate that there is some other mechanism at play than network size (which has been the core operating thesis of "AI")
"Studies on animal minds suggest consciousness is not computation"
https://iai.tv/articles/studies-on-animal-minds-suggest-consciousness-is-not-computation-auid-3535?_auid=2020

2026-03-31 08:49:17 UTC

“[2603.21687] MIRAGE: The Illusion of Visual Understanding”
https://arxiv.org/abs/2603.21687
> Frontier models readily generate detailed image descriptions and elaborate reasoning traces, including pathology-biased clinical findings, for images never provided
And
> Second, without any image input, models also attain strikingly high scores across general and medical multimodal benchmarks, bringing into question their utility and design

2026-03-31 00:34:31 UTC

It's clearly going to take the tech industry a while to understand that the "AI" brand is well and truly toxic
If you have a software or feature, sell it on what it does for people. If it's actually useful, then calling it "AI" will just make people assume it's yet another money-grabbing con.
Especially if you aren't using an LLM or diffusion model and instead are just using a non-generative ML model. Why willingly adopt all that negative baggage when you don't have to?

2026-03-30 14:09:21 UTC

- reply

"Entitled" is indeed an apt description of quite a few people I've encountered in tech.

2026-03-30 13:55:26 UTC

- reply

Yeah, I've had that experience in the past as well.

2026-03-30 13:37:51 UTC

- reply

I agree that much of the discourse has given many tech workers odd ideas about how intellectual property works
And it's not depriving people of revenue that's the distinction. RSS is offered by the publisher. Even before ad blockers, web ads have always been offered with the understanding that not every visit will result paying ads (ad inventory is a thing)
LLMs and diffusion models promise to replace creators entirely using their own work. That's an important distinction

2026-03-30 13:32:51 UTC

- reply

That's a a solid article. i don't think I'd read that before. Thanks!

2026-03-30 13:31:37 UTC

- reply

Hah! Thanks 🙂

2026-03-30 12:53:10 UTC

- reply

Yeah, naming names in this context usually leads to a bunch of harassment for everybody involved. I was vague for a reason.

2026-03-30 09:36:26 UTC

- reply

Yeah, that’s definitely a possibility

2026-03-30 09:21:57 UTC

So, serious questions. I keep seeing people in tech talk as if they don’t see the difference between indexing media to aid in the discovery of said media and indexing it all to replace it and deprive the creators of revenue.
Is it really not obvious to people in tech why people in the arts and creative industries might be fine with the former but extraordinarily furious about the latter? Or that the legal framework surrounding the two might actually be different?

2026-03-29 16:06:14 UTC

Finally feeling caught up after losing access to the good stuff* for 24 hours.
(I'm such an internet addict. Not so much social media but having All The Things on tap: reading, videos, audio, music, whatever strikes my clearly massively neuroatypical mind at each moment)
* fibre-optic broadband

2026-03-29 11:34:37 UTC

- reply

Oh yeah. Absolutely

2026-03-29 11:34:09 UTC

- reply

Yeah. Turns out they very definitely can 🤦🏻‍♂️

2026-03-29 11:33:17 UTC

- reply

My pleasure.

2026-03-29 11:32:59 UTC

TFW tech boosters co-opt the language of critical analysis and advocacy to argue the only way to fight “bad AI” is to make “good AI” not acknowledging that it doesn’t actually solve the problems they themselves copy-pasted—sorry, I mean cited—from those who have been talking about it for years
(This is a generalisation, not a sub-post, a distinction that matters. I’ve seen several guys make this same argument in this same way over the past month)

2026-03-29 11:01:59 UTC

“Grammarly shows how prototyping turned into an excuse for not thinking”
https://productpicnic.beehiiv.com/p/grammarly-shows-how-prototyping-turned-into-an-excuse-for-not-thinking
> But in the moment it actually mattered, critical thinking was swept away by the false urgency of pushing things out.

2026-03-29 10:57:48 UTC

RE: https://mastodon.social/@iris_meredith/116311817974687217
> In short, if friction is neglected, morale quickly ends up in the toilet and no progress to speak of can be made.
The whole piece is very good
#note1zzx…trat

2026-03-28 15:36:36 UTC

- reply

Touchscreen typing having also been retroactively devolved into the worst input method ever with degraded autocomplete doesn’t help either. Those blow-suck accessibility straw interfaces probably have a higher accuracy rate at this point than the iOS keyboard.

2026-03-28 15:31:56 UTC

Main internet’s busted so here I a. 5Ging it like a protozoa expressing itself from the primordial ooze. Yuk.

2026-03-28 15:22:40 UTC

“You Do Not, In Fact, Have to Hand It to Them”
https://2ndbreakfast.audreywatters.com/you-do-not-in-fact-have-to-hand-it-to-them/
> Their technology, their politics, their vision of the future -- none of these are what we want. More importantly, none of this is inevitable.

2026-03-28 15:11:49 UTC

“On The Enshittification of Audre Lorde: "The Master's Tools" in Tech Discourse”
https://tarakiyee.com/on-the-enshittification-of-audre-lorde-the-masters-tools-in-tech-discourse/
> Read against the speech itself, Lorde's argument seems less concerned with whether antitrust law can break up monopolies and more with whose knowledge counts, who gets to define the problem, and what gets systematically erased when liberation movements reproduce the exclusions of the systems they are opposing.

2026-03-28 11:40:42 UTC

That feeling when you get a “too many requests” error on Github for clicking a single link on a single page.