Notices by Baldur Bjarnason (baldur@toot.cafe), page 2

“Google developers significantly misstate carbon emissions of proposed UK datacentres | Google | The Guardian”

https://www.theguardian.com/technology/2026/may/09/google-developers-significantly-misstate-carbon-emissions-of-proposed-uk-datacentres

> Emissions understated by factor of five

Why it's a mistake to take claimed environmental impact of data centres at face value

“Real signals or artificial stereotypes? - by Adam Kucharski”

https://kucharski.substack.com/p/real-signals-or-artificial-stereotypes

> Despite the responses being identical for the UK and US, Copilot produced a rich, detailed summary of how US and UK respondents differed.

I know too many people who are using LLMs for exactly this kind of data analysis and refuse to listen to me when I try to warn them

This is a long-standing characteristic of the field and it explains a lot of what’s been going on in tech. If you can’t even empathise with people doing the same job as you—that share your circumstances—what are the chances of you caring about the end user or the effects of your work on society?

The class solidarity argument has been a part of “AI” discourse in many fields for a while now. It doesn’t settle the debate but it’s a rationale many will acknowledge even if they disagree

Except in coding. Mention it to software devs and most will look at you as if you just spoke in tongues

I'm worrying that this is starting to matter a lot as it looks like expertise is an important qualifier limiting the harm that comes from using LLMs for coding. If this is a general observation and not just a reflection of my poor career choices then it seems likely that the client-side web specifically will be hit harder by the code "slop-apocalypse" than other software.

Basically, I know a lot of web devs with 10+ years of experience whose knowledge about any given topic in their field is at about the level of a recent graduate. Most of what they know is hearsay and superstition and most of what they do is play around with trends

And, again this is in my experience and I may just have been quite unlucky, this is more common in web dev than other parts of the software industry.

One issue I rarely see mentioned are the sharp differences between expertise progression in general web dev, online web dev subcommunities, and the rest of the software dev community.

Namely, in my experience, it's very common in general web dev for people not to have the expertise you'd expect based on their seniority that you see in the people who have the dedication and interest to participate in discussions on specialities within their field.

“Our Survey on Creativity, Writing, and Reading in the Age of AI”

https://ellipsus.com/blog/survey-on-writing-and-ai

I think few people in tech really get how truly unpopular "AI" is shaping up to be

5. Company announces that they are working on the next gen “this time we really mean it” tool

6. “OMG! They must be intentionally degrading their old tool to sell the new one!” (No, what you have is a stochastic piece of shit that sometimes works out of luck and you suffer from a mind whose only mode of thinking is a whatever confirms your existing biases.)

7. The co releases the next gen and the cycle starts again

2/2

The seven step “OMG! My AI tool is suddenly degrading” cycle:

1. An LLM tool is released. The company claims it’s a miracle that will change everything

2. Early adopters try it on straightforward tasks, find it seems to work, though they don’t really check it. “OMG! It’s a miracle that will change everything!”

3. Adopters apply it to an increasing variety of tasks. Some work; some don’t because the tool is still fundamentally random

4. Novelty wears off and error reports start

1/2

RE: https://mastodon.social/@iris_meredith/116395821486826615

This is very much a tangent inspired by Iris's article but...

The reading dichotomy Iris presents (grammatical versus Bayesian sentiment analysis) maps pretty well to McLuhan's hot and cold media, which in turn maps to Tarkovsky's montage vs time pressure, Tolstoy's art formed of associations with other art vs that formed of experience and emotion...

This spectrum of "reading" (despite being very annoying McLuhan was right to avoid a binary dichotomy and instead use the metaphor of a temperature range) have existed for a long long time but have only really become unbalanced in a problematic way over the past few decades

This is quite specifically what Neil Postman was warning us about in "Amusing Ourselves to Death" in 1985 and his warning ran up against the same wall that blocked other predictions such as that of global climate change:

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

I keep encountering, online and off, people who believe if an unethical tech becomes functional enough, that will make all the naysayers abandon their concerns and start using

That is, if LLMs become good enough then surely we will look past the deception, violations, abuse, and extremist politics?

The whole "my magic genie can break open any software in the world" thing is so suspect

And now it has media types imagining it can not only find vulnerabilities, but vibe code a workable exploit, cut through multiple layers of security and then hop onto an air gapped system to take down modern society.

I know most software is shit, but the infrastructure systems modern society relies on are, for the most part, designed with the assumption that they have undiscovered vulnerabilities

And the rest, the shit systems that aren't core infrastructure, are packed with KNOWN vulnerabilities.

Much of the web is an all you can eat software exploit buffet. People should be much more worried about malicious people finding ways to automate the application of known exploits than about systems finding hard-to-exploit vulnerabilities in otherwise hardened software.

(Also don't use Windows for anything that needs to be safe. Just my 2 kroner.)

RE: https://toot.cafe/@baldur/116370347475754359

Am I going to have to repost this thread every day now? Please be a little bit more sceptical about claims along the lines of “the secret thingy I’m trying to turn into a billion dollar payday is super powerful and almost magical, be afraid, but also give me money”

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 the lying liars will be people you’d trust in normal times because the first person they fool is themselves

Be extra sceptical.

(Talking trash about Thatcher on social media has a very utilitarian purpose: Thatcher fans are some of the most annoying people on the internet and the best way to improve your replies is by having as few of them around you as is possible.)

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