[LLMs] Seeing the Linux kernel’s LLM policy being summarised as ‘puts the responsibility on humans’ or some such, and… so? First of all that’s what every LLM user tells themself whether it’s true or not. Actually making them take responsibility is harder, because the task of enforcing that […]

[uspol] american conservatives: "we want our leaders to be pro-life!" the pope: "war is bad" conservatives: "no not like that"

Go easy on Claude users right now. They're all mourning the loss of their imaginary friends. https://github.com/anthropics/claude-code/issues/45596

"#mythos discovers 27-year-old bug" is intentionally conflating length of existence of a bug with difficulty of finding it. There is no such correlation. If software projects and companies performed regular, ongoing, in-depth code audits over and over and missed it, sure, then age would be […]

I think about this a lot

There is no ceasefire in Lebanon. Israel has been bombing Lebanon non-stop since last night. If anything Israel is expanding its war on Lebanon. You are being lied to

RE: https://toot.cafe/@slightlyoff/116350637495317031 it’s always amusing to see a microsoft employee complaining about apple’s anticompetitive practices

[LLM security self-own] Honestly I have zero sympathy for anyone running OpenClaw because they absolutely all should have known better, but maybe this latest gaping security hole will finally teach some people some sense […]

Successful surgery on my BGP core for AS201379 done. Got a third edge-router online, reorganized local-preferences and optimizing packet flows for efficiency and cost. Monitoring with Grafana is perfectly fine. RIPE Atlas measurement looks amazing for […] [Original post on burningboard.net]

By all accounts the Claude codebase is catastrophically shit (and built using Claude itself, obviously) so I absolutely do not want to hear from anyone who's impressed by it, thinks it's impossible to justify not using it, etc.

People keep assuring me that LLMs writing code is a revolution, that as long as we maintain sound engineering practices and tight code review they're actually extruding code fit for purpose in a fraction of the time it would take a human. And every damned time, _every damned time_ any of that […]

“Let’s agree to disagree” No

We regularly watch TV with the subtitles turned on, and sometimes the quality is ... underwhelming. My favourite bad example is an episode of Frasier in which a character uses the phrase 'bête noire' and it gets subtitled as '[speaks foreign language]'. There's also a common problem when we […]

another nonsensical ai take in the guardian https://www.theguardian.com/technology/2026/mar/27/number-of-ai-chatbots-ignoring-human-instructions-increasing-study-says obviously it's a concern when an agent does damage, especially when people insist on giving them elevated permissions but it's […]

Man, LLMs really drive home the lesson that people differ *greatly* in how much the care whether the statements they make are true.

RE: https://mastodon.social/@jessie/116301692264356354 i must not sign up for cornish lessons i absolutely will not have any free time or mental energy for this and if i did i ought to focus on finnish or even swedish and yet.

[gbpol, tory freaks, education] birbalsingh is regularly referred to as ‘britain’s strictest headteacher’ but as far as i can see, based on her public persona, she’s completely ineffectual her entire management style seems to consist of learning that a child […] [Original post on hachyderm.io]

New MacOS version prompts me to enable AI so I can have ‘custom emoji’, then warns me that opening Emacs might not be safe. I’m not sure the priorities are correct here.

RE: https://infosec.exchange/@BleepingComputer/116295860536457673 whatsapp adds AI-powered message replies so you can automate the boring task of talking to your friends and free up more time to work

[transphobia, rowling] ah i see we’re doing jk rowling discourse. again. pro tip: if you actually do hate trans people then just donate to transphobic organisations directly and cut out the middle man. otherwise, stop giving money to transphobes that they will immediately spend on doing harm […]

setting the 'days since an infuriating bug turned out to be the fault of me, nine months ago' counter back to 0

> i fucking LOVE REVIEWING CODE > I LOVE IT > WHEN PEOPLE I KNOW HAVE IDEAS > AND THEN THEY DO SHIT THAT IS COOL > AND THEN WE TALK ABOUT THOSE IDEAS TOGETHER > BECAUSE WE HAD THEM > IN OUR OWN MINDS

just saw someone (on linkedin, obviously) using the phrase 'deploy a designer resource' and i do not understand why people can't just talk normally about other people

OpenAI isn't getting the money they want from investors, is killing the Sora video generating app, and is set to almost double their human workforce, is billions in the red again this year, while banks and other large companies are scaling back on data center investments. Can we NOW admit that […]

i'm at a loss of words after reading a paper about _reformatting code_ using an ML model that has a measured statistical quantity A_c which says how often the reformatted code behaves the same as the original the "ideal" (their choice of words) case is 64.2%

"Programming advice like 'Don't Repeat Yourself' is now obsolete. Much easier to get AI to update all instances of the repeated code than to avoid repetition." This is what I get for reading Linkedin, I suppose. I brought it on myself.

https://www.themonastery.org/blog/why-some-catholics-want-peter-thiel-burned-at-the-stake i know the catholic church has its issues but ... let's hear them out on this

I mean what do I even say? Despite all my issues with Hezbollah, it is a fact that they did respect the ceasefire and even the BBC acknowledged that. Israel broke the ceasefire over 15,000 times in 13 months (yes, literally 15k) But the ceasefire apparently ended when Hezbollah broke it one time

New blog post: Reading (January–February 2026) https://eskola.uk/2026/reading-january-february/

Getting adverts on linkedin for weapons manufacturers, for some reason. Reporting them all as 'threats or violence', not that it'll do any good.

blocked by that bluesky dude for pointing out that 'we forgot to make an important announcement about the status of the company for an entire eleven months' is not a persuasive justification, and if true just makes them all look like they're terrible at their jobs

"We asked experts about the most responsible ways to use AI tools – here’s what they said" funnily enough, none of them said not to use it at all 'responsibly' apparently means 'don't use it for certain things' or 'be transparent' increasingly difficult to trust the guardian these days

since I see a lot of gloom re Astral x OpenAI: ⁃ Even if Astral actively (malice) or passively (neglect) ruins uv, Python packaging is 1000x better today than 2 yrs ago. ⁃ Ruff is nice but only nice to have—ty is barely usable. We'll be fine if they vanish today. ⁃ Channelling Buddhism: it’s OK […]

This is how you judge the Starmer government. Weapons more important than child health. No morals. Just panic. Some of the world’s poorest countries to lose UK aid due to 56% budget cut | Global development | The Guardian […]

Oh look the VC-backed Python tooling company that everyone loves has been acquired. i.e., the concern everyone had right from the beginning about a VC-backed firm, because there's obviously no money in making a faster linter. Good news everybody, your syntax checker is now implicated in war crimes.

can't believe the british government is so completely dead set on never allowing independence for scotland, wales, or the north of ireland, yet the bbc has just announced the border between england and scotland is a 'coast path' https://www.bbc.co.uk/news/articles/cy0dxexdd8xo

Britain has “failed in its fundamental obligation to prevent genocide, has been complicit in atrocity crimes, and in some instances has even been an active participant in these crimes”. #Gaza #UK #IsraelWarCrimes #Corbyn […] [Original post on kafeneio.social]

RE: https://hachyderm.io/@SnoopJ/116206175581735706 I had missed that dario amodei made a NEW stupid prediction of exactly the same sort in January of this year: https://finance.yahoo.com/news/anthropic-ceo-predicts-ai-models-233113047.html So, set your "stupid prediction doesn't come true" […]

I've a better idea. Unilateral Nuclear Disarmament. Davey advocates what the Western consensus deplored in Iran. Double standards UK must build own nuclear missiles to end US reliance, says Ed Davey - BBC News https://www.bbc.co.uk/news/articles/cy0dz1k0rr4o #nucleardisarmament

https://www.axios.com/2026/03/14/israel-lebanon-war-peace-hezbollah-france “The French government has drafted a proposal to end the war in Lebanon that would require the Lebanese government to take the unprecedented step of recognizing Israel” So: Israel can do whatever it wants, and in return […]

Fastmail @fastmail has been operating on bare metal for 25 years. Their planning and (over-)provisioning strategy reminds me of Wikimedia. Situations where cloud renting is cost effective for medium or large companies are vanishingly rare, and always have been. As with so many big tech patterns […]

I've never liked the autocomplete suggestions — I turned off even the individual-word suggestions in iOS years back, and the phrase suggestions are much more obnoxious than that. Apparently they're even worse than I'd realised, though […]

RE: https://mastodon.social/@mcc/116217146358807272 ‘why do people hate “AI”? it is a mystery’

RE: https://mastodon.bsd.cafe/@stefano/116215994258493990 LLMs don’t understand FreeBSD? FreeBSD sounds better and better.

Fingerprint scanning on Linux seems very unreliable. I have it set up, and it can definitely recognise a fingerprint and authenticate based on that, but there seem to be far more false negatives than I’m used to. Can’t tell if this is the fault of #fprintd or if the #thinkpad fingerprint reader […]

"There were some ethical concerns about using AI but we decided to set those aside" oh that's all right then

[gbpol] This has taken far too long, and frankly doesn't go far enough in terms of Lords reform, but it's better than nothing. https://www.theguardian.com/politics/2026/mar/10/hereditary-peers-to-lose-their-seats-in-the-house-of-lords

I don’t know why I ever try to be good at my job when I could just build stupid bullshit that nobody needs and that makes the world a worse place and the worst people in the world would line up to throw stupid amounts of money at me […]

setting the 'days since a bug was caused by vibecoding' counter back to 0

love to go to a programming meetup and be told that programming is obsolete and i should just get clyde to do it all instead 🙃