Baldur Bjarnason

If the pope released a statement that the world was ending next Tuesday with the rebirth of the Messiah, the job of a reporter isn't to schedule an interview with Jesus H. Christ reborn but to figure out what's wrong with the pope Media accepting "AI" millenarianism as fact is genuinely disturbing

A decade ago during a project, I was griping about the trouble friends of mine had been having with their landlord, who was controlling the thermostat remotely to save money (rent was inclusive of bills). This was a shit thing to do when you don't know a family's schedules or even the local weather… But when I was venting about this, the other devs got defensive and angry. Turns out they were both landlords who did exactly the same thing to their tenants

“Bad Vibes Coding – Spectre Collie”

https://spectrecollie.com/2026/09/05/bad-vibes-coding/

My take on generative AI has pretty rapidly gone from a tentative, “it sucks and I hate it, but after the bubble pops, there might be a few very specific use cases for it, like with programming,” to a vehemently zero-tolerance stance. It’s all garbage, burn it down.

All this engineering and all this money and both tech and software just get less and less interesting by the day. All this energy; so little imagination.

As many others have pointed out, one way to think about what’s been happening to tech—software, especially—is that the entire industry is being “Boeinged” (or Jack Welched). And, as with an airplane, a faulty part can sit in a system for a decade before it snaps and brings the whole thing down

So the Netlify bill for the surprise bandwidth overage caused by a bunch of badly written services and scripts downloading and re-downloading the podcast episodes comes down to $193 USD The bunny.net charge for similar bandwidth would have been about $13 USD 😬😬😬

Your semi-regular reminder that given the size and extent of the “AI” bubble it’s an absolute certainty that fraud is involved in it somewhere. We don’t know the scale or location of the fraud but you simply do not get a trillion-dollar bubble without fraud, both financial and in information/discourse And all that fraud is going to be lined up supporting “AI”. Calibrate your scepticism accordingly.

Think about outsourcing. There was no "after the outsourcing bubble" for most western factory workers. There was just immiseration and ruined towns. There was no "after the 2007 housing bubble". We still have housing bubbles pretty much everywhere you look. The finance markets are even less regulated. Fraud was basically legalised. So, I'm sceptical about the notion that the software industry will "revert" after the bubble.

Just heard that a CEO I know (or used to know back in the day) is using a ghostwriter for their linkedin posts. Which I honestly kinda respect. Still fake as hell but at least a step up from slopping the posts yourself. Outsource the slopping to somebody else, FTW.

“UK AI medical transcription is dangerous and doesn’t work – Pivot to AI”

https://pivot-to-ai.com/2026/09/03/uk-ai-medical-transcription-is-dangerous-and-doesnt-work/

The vendors always claim that patients trying to get an appointment out of the AI can totally speak to a human any time. This turns out not to be true.

Generative models have no place in healthcare.

I admittedly i don’t have that many coding client projects left these days but in the ones I’ve got left, most of the coders I work with have adopted LLMs For what it’s worth, they do seem to understand much less, work much slower towards the project’s actual goals, with much more code, and at a lower quality While asking sillier questions But they think they’re working so much faster because the tools find the occasional bug and generates so much code 🤷🏻‍♂️

“I Don’t Want to Interact With Stochastic Parrots”

https://ploum.net/2026-09-09-ai_policy.html

My AI policy is "I don’t want any kind of interaction with that thing." And this should be all you need to know.

This.

“What kind of Software should we fund?” https://tante.cc/2026/09/17/what-kind-of-software-should-we-fund/ Separately from what I’m responding to, Tante, though, does a good job of gleaning useful observations from the original post. Read this; don't read the original it's responding to

As far as slop indicators go, variations on “let me break that down for you” have to be somewhere in the top ten Also, do people not realise how condescending and insulting this phrasing is in a reply to a support request? Do they even read their own responses?

“The “Nonexistent” Research on AI’s Benefits for Education”

https://www.insidehighered.com/news/tech-innovation/artificial-intelligence/2026/09/14/nonexistent-research-ais-benefits-education

But according to preliminary findings he published last month, several of the existing peer-reviewed studies that say AI is an effective learning tool are based on flawed methodologies and analyses that rely on misapplied, miscalculated or misinterpreted statistics.

“The death of web development education – Rescuing a field from disappearing”

https://molily.de/web-dev-education/

Rescuing a field from disappearing

“Why office workers are turning against AI”

https://www.bloodinthemachine.com/p/why-office-workers-are-turning-against

It’s also because bosses are much less accountable for their AI use. They may use it to write emails and company memos or to experiment with task automation, but they’re not going to be disciplined or fired if there are errors or inconsistencies in that stuff.

“Anthropic IPO filing: the disaster in numbers – Pivot to AI”

https://pivot-to-ai.com/2026/10/01/anthropic-ipo-filing-the-disaster-in-numbers/

Anthropic warns how many of its largest clients are not locked into long-term contracts and could drop Anthropic in a moment.

People love RSS for the same reasons hikers love a good path. It isn't the packed dirt or gravel they're fond of, but the view it takes them through.

"'Software issue to blame' for UK flight disruption" https://www.theguardian.com/business/live/2026/sep/18/japan-raises-interest-rates-central-bankers-fight-inflation-gb-retail-sales-rise-latest-news-updates?CMP=share_btn_url&page=with%3Ablock-6aad30998f085ee1fe6a0621#block-6aad30998f085ee1fe6a0621 Of course it was a software defect. We're going to be seeing a lot more shit like this happen over the next few years as people discover that the software industry is being unravelled not "revolutionised"

That EU nations are aggressively turning on and threatening each other at every opportunity and refuse to come to each others aid during crises, right as Icelanders are preparing to vote on a referendum on whether to reopen EU application negotiations, sure is great timing /s

“Poor genie keeps trying to crawl back into the bottle. Everybody hates him.”

Ceci n'est pas une pipe

“Anthropic watermarks AI output — and the AI bros yell – Pivot to AI” https://pivot-to-ai.com/2026/08/17/anthropic-watermarks-ai-output-and-the-ai-bros-yell/ Watching all of this unfold has been kind of funny.

“DHH Nazism Funded by 1Password VP Who Wrote “Honest Security” | flyingpenguin” https://www.flyingpenguin.com/dhh-nazism-funded-by-1password-vp-who-wrote-honest-security/ It's certainly starting to look like much of tech is run or financed by Nazis and Nazi sympathisers. As in, out and out "will to power" fascists

That feeling when you see people get away with only criticising the perceived overreach of those against "AI" but never those promoting it; somehow managing to look "impartial" while making money advising companies on "AI" You just know they'll come out of the other end of this bubble squeaky clean One of the downsides of having known a lot of consultants is an intimate familiarity with polished and accomplished dishonesty.

@davidgerard@circumstances.run I don’t know what’s going on with him but Adam Conover comes across these days as somebody in a post-divorce mid-life crisis with an associated cash crunch

RE: https://mastodon.social/@freakonometrics/117054108271195109

“The Gargantuan Lie That is Collapsing The World’s Climate”

https://www.currentaffairs.org/news/the-massive-climate-lie-that-will-destroy-human-civilization

This massive lie can best be summed up as “when it comes to climate change... we have more time.”

This is actually the main reason I'm so angry about the "AI" bubble. All this enormous effort could have been spent combatting the climate crisis and the opportunity cost of your fucking chatbot is going to be the very climate we live in

Imagine what we could have done with those 3 trillion dollars (or so)?

“I Hate Adobe” https://aftermath.site/i-hate-adobe/ Yeah. Same.

I find the defensive panic of US ‘AI’ enthusiasts in response to the use of Chinese models kind of comforting “Look, if you actually do the numbers, the Chinese models aren’t cheaper!” You don’t say. Really? Do you mean to say that they aren’t going to make this whole house of cards sustainable after the bubble pops?

Because people are already being annoying about this: it’s obviously not notable when somebody who dislikes using LLM tools and worries about the impact of “AI” complains and ponders leaving the industry. It is notable when people who think the tools are useful and have no moral objection to them also express the desire to leave industry. Obviously not as common but also obviously more notable.

I see English-language media is covering the upcoming referendum in Iceland—“do you want to reopen talks to join the EU"—without much context. This is understandable as Icelanders generally deliberately withhold that context to manage Iceland's image So...

@dngrs@chaos.social though lead poisoning has broader support among the populace than “AI”

Almost forgot to add the photos themselves. These are from a new path that was made in the reforestation area to the north of the town to a small grove called Ísakslundur that wasn't really accessible before

I framed my book on generative models in terms of risks for a reason. Whatever you can say about their "productive value", their nondeterminism, manipulative UI, statistical mediocrity, and tendency towards unpredictable errors mean the risks—even if you're an asshole who disregards societal risks—outweigh the still largely hypothetical benefits

Unreliable models whose output is massaged into place by overseas labour will just suit them fine, even if the result is largely broken. See outsourcing I'm guessing there won't be an "after the bubble", at least not for software dev . This might well be the future of coding in perpetuity, just at some point it'll be even shittier This can only be fixed on a political level. A bubble pop will give the investor class an opportunity to solidify the system into place, much like after 2007

So, the post-bubble equilibrium models will either be increasingly stagnant, built using stale training data, extended and tweaked as best as people can, but get stranger and stranger the more they are fine-tuned Or the models will be something completely different shaped by the new economics But whether the tech itself is effective or not ultimately isn't the point which, for the the capital class is less local labour.

@alda@topspicy.social 😆

“Support Pivot to AI — video previews, patron name reads, T-shirts! – Pivot to AI” https://pivot-to-ai.com/2026/07/28/support-pivot-to-ai-video-previews-patron-name-reads-t-shirts/

Weekend confession time: One of the reasons why I’m extra sceptical of the various enthusiastic “AI” critics who manage somehow to sound vociferously critical of “AI” without actually criticising the tech itself and somehow remain eternally ignorant of the many contributions women (and specifically women) have made to the field of “AI” and “AI” criticism, is I keep asking myself: Which these verbiage lovers will turn out to be the Matt Taibbi of “AI” critics?

@alda@topspicy.social @RandamuMaki@mstdn.social Too true.

@Dangerous_beans@aus.social Too true.

One core problem is that the tech is volatile. You can get very different models from training run to run even using the same data. Current models can do what they do is because they keep throwing money at runs and training data Because of this volatility, this isn't a tech that works like a foundation. It has lottery dynamics