Baldur Bjarnason

😄

The diff is that LLMs are variable on every go, every time they’re used. It’d be like if Gutenberg’s printer, after the type had been set, varied in how long it took to print a sheet from a second to ten minutes. Or if you didn’t know whether your model T was going to run at 5mph or 50 each trip

Unpopular opinion: the enshittification of the only thing that takes git from "so actively user-hostile it causes significant barriers to actually doing anything" to "tedious and annoying but grudgingly tolerable, mostly" can only hasten the day someone makes a better DVCS and this is Good, Actually

People keep describing “AI” as “low effort” but that’s not accurate. It’s high effort on the back end (training, etc) and variable effort on the front.

Sometimes it saves work. Sometimes it’s more work.

This doesn’t really have any direct parallels in industrial or tech history

Given that political history is repeating itself (fascism and authoritarianism) in a collapsing environment expecting technological history to repeat itself is an incredibly unsafe assumption.

oh my god RIGHT????

if a person talked to me the way a lot of these LLM transcripts go I would back away and then probably mute them, IT IS SO UNCOMFORTABLE

Dev posts on sporadic personal blogs that don’t involve “AI” and aren’t written or co-written by an LLM (you can tell from the sudden change in writing style) are becoming rarer and rarer.

I follow hundreds of blogs in my feed reader, mostly very sporadically updated personal blogs by software devs and every day there is now a new blog post from an "AI" convert.

And always it works great at tasks outside of their expertise and especially great at the stuff they don't check

A rusty abandoned farmhouse surrounded by overgrown plants.

At the beginning of the 19th century, there were 13 farms in the area around Stóri-Hólmur, known as Leiran. Leiran relied mostly on fishing for survival. There was even a school there at one point.

#Iceland #photography #naturephotography #nature #landscape #abandoned #farm #plants #sea

“I tried coding with AI, I became lazy and stupid - Thomasorus”

https://thomasorus.com/i-tried-coding-with-ai-i-became-lazy-and-stupid

> The LLM had given me shitty code, made me ignorant about my own code base, and too lazy to try to fix it myself.

með viðbjóðslegt transfóbískt skilti, eins nálægt Trans Íslandi og hann komst. Viðkomandi málaði sig í kjölfarið sem fórnarlamb þegar fólk í kringum hann fékk nóg og bað hann um að fara. Löggan aðhafðist ekkert. Þvert á móti tóku þau við kæru frá HONUM fyrir að skiltið var skemmt. Þetta gengur ekki.

Forseti Trans Ísland: „Megum ekki slaka á þótt við séum að mælast ofarlega“ - RÚV.is Ísland er í fyrsta sæti á réttindakorti trans fólks í Evrópu og Mið-Asíu. Forseti Trans Íslands fagnar því en segir Ísland þó ekki mega slá slöku við í réttindabaráttunni.

Ísland má eiga það að vera framarlega í alls kyns mannréttindabaráttu, en það má ekki gleymast að þó að við séum „best” að þá er það ekki samasem merki við að fullu jafnrétti sé náð, einfaldlega að önnur ríki standa sig verr.
Í gær fékk einstaklingur úr „samtökunum 22” að ganga óáreittur í göngunni-

Sarah Polley: ‘It took me years to see how responsible Terry Gilliam was for my terror’ As a child, the Canadian actor-director starred in the former Python’s The Adventures of Baron Munchausen. Here she recalls how she feared for her life on his chaotic film set

Excerpt

Sex is getting scrubbed from the internet, but a billionaire can sell you AI nudes Grok’s nonconsensual deepfakes show how hollow the online safety push can be.

Sex is getting scrubbed from the internet, but a billionaire can sell you AI nudes

a chart showing rising dc power consumption but flatlining renewable growth levels

Nice FT citing Ireland, where data centre power demand has massively outpaced clean power growth - meaning increased burning of fossil fuels

You can't rely on an assumption of infinite, exponential, unstoppable renewable energy to excuse data centre expansionism.

www.ft.com/content/0f61...

The ROI of UX is so massive that no one will believe you User research makes the most difference when findings challenge closely-held assumptions. But the truth is often unpopular.

The ROI of UX is the value of your entire company, because user research is the work that determines what problem is most valuable to solve in the first place.

Trouble is, people won't believe you if you tell them this.

So the second job on top of that job is to convince people to trust us.

Somebody on LinkedIn said what we're all thinking.

There’s a vibe rn where many of us in the UK are stopping to take stock of what went wrong over the past 8 years with trans ppl, and I do think it’s important to reconsider tactics, but at the end of the day there was only one decisive factor in all of this: money.

The main newsletter would still get both the depressing stuff and the items that are attempting to be more uplifting, but those would be separate entries instead of bundled at the end of the newsletter as I do currently. That way I could set up a separate newsletter with just the "uplifting" entries

My social media and newsletter is a mix of depressing tech analysis and photos of cats and similar feelgood items

A half-baked idea I have is to spin out the feelgood stuff into a separate newsletter, coupled with writing that attempts to be more positive but I have no idea if there'd be interest?

It’s the practice that matters the most, because with the right practice you can pick up the reading as needed

The thing is, the various fields bundled under the “humanities” label are practices, not reading lists. You can easily graduate from comp. lit. for example without having read much of the conventional “canon” and that’s just as valid as somebody who has done the entirety of Shakespeare

This is particularly laughable for Perplexity who has 60% gross margins but over 15 million free users. Their real margins are probably negative 30%

Yeah man our gross margins are great as long as you don't include our largest costs

Just read that again: AI tech bosses are choosing not to have children because they don't think the world will be around much longer.
WHY are we inviting this monstrous tech into all levels of education? Why are we blithely going along with Big Tech as it tells us we cannot do without AI?
WHY?

Like the writer, I very much dislike Candide (except in Laxnes's Icelandic translation which is unfaithful and highly entertaining), but that's very much not why you read it in comp. lit. or philosophy.

This is "humanities" cosplay, where techies take a reading list across a broad range of fields and proceed to plow through them as recreational reading at a pace that rivals a tourist "doing" the US or Europe in a single week …

“Humanities Crash Course Week 32: Voltaire”

https://jarango.com/2025/08/10/humanities-crash-course-week-32/

This is too illustrative of an example to pass up. This specific reading of Candide is an entirely valid personal take, but it is specifically not any flavour of "the humanities" …

Karen Hao on AI tech bosses: ‘Many choose not to have children because they don’t think the world is going to be around much longer’ The author of Empire of AI: Inside the Reckless Race for Total Domination discusses the cost of Big Tech’s huge investment in technologies that may do more harm than good

Karen Hao on AI tech bosses: ‘Many choose not to have children because they don’t think the world is going to be around much longer’

The ER story https://www.ruv.is/frettir/innlent/2025-08-10-ovanalega-god-stada-a-bradamottokunni-eftir-laugardagskvold-450550