Laurie Hedge

@JessTheUnstill@infosec.exchange truly a dream job!

@JessTheUnstill@infosec.exchange I'd like a "diagnose plant" cantrip that tells me whether my plant is dying because I'm over watering it or under watering it or using the wrong kind of water or too little sun or too much sun or not enough plant food or too much plant food or the wrong kind of plant food or the soil ph is wrong or the roots are tangled because they all kinda look the same

A few years back I set myself the challenge of learning enough Old English to have a go at translating Beowulf. I started my attempt at a translation back in January, and am about a third of the way through the first draft. I had particular aims for this, specifically to do a colloquial translation with a focus on readability.

Anyway I really depressed myself tonight by giving the free version of ChatGPT the first 20 lines or so and asking it to do a translation in a colloquial and readable style, and in less than 30 seconds it had honestly done a better job of it than me..

I really wish I could say it was terrible slop, but what's so upsetting is that it isn't. I'm pretty convinced that if I removed the tells from its and my versions (e.g. removed the em dashes it uses, ensured mine is free of typos etc.) a human reader wouldn't be able to guess which was which.

@dragonfi@social.jsteuernagel.de Well... I'm not sure if it helps but there is a fully scalable PDF version here (I only rasterised it to PNG for posting but I made it as SVG) https://nextcloud.lauriehedge.uk/index.php/s/2Xj6gHHtx9Xm98i

@atomwitch@tech.lgbt Yay! Then I consider it time well spent :)

Don't ask how many hours of my one and only life on this earth I spent making this.

Brought to you by me remembering again that Ikea is a fully paid up promoter level member of the Khronos group, the answer to the question: what if Ikea made the Vulkan tutorials?

In case it becomes relevant again, just a reminder that Andy Burnham voted consistently for the Iraq War, consistently against an investigation into the Iraq War, and for all Blair's surveillance state shit like ID cards.

Breaking: Starmer and Badenoch cite The Black Knight from Monty Python as their political role model

Since the rise of AI coding, I've seen non-ascii characters in source files become completely normalised in English language codebases, primarily in comments but also in string literals (I've not seen it in identifiers yet). I wonder how many compiler bugs this has surfaced? Or maybe non-English codebases already flushed out all the UTF-8 bugs?

Hard not to feel left out with all the UK folks posting about going to the polls today when there's no election in my area. I would have enjoyed the opportunity to give the government a good kicking and to do my bit to keep Reform out. I hope everyone who does have an election today makes the most of it!

It's great how we have a terrorism threat level system that accurately predicts what the likelihood of a terrorist attack yesterday was.

I don't want to be churlish about this because I do think the government's reforms to leasehold are a lot better than nothing, but seeing them try to explain how their manifesto commitment to "finally bring the feudal leasehold system to an end" didn't actually mean that they intended to abolish the leasehold system is just infuriating.

Scene 1
(Laurie, coffee mug in hand, is sitting in front of his computer, a clock widget on the desktop shows it is 9:30 on a Monday morning)
Laurie: (talking quietly to himself, slowly raising mug to take a sip) Oh look, a PR came in over the weekend, well I guess I'll start my morning by getting that reviewed and then I'll
(camera zooms into the PR metrics, +12,000 -2, camera shake effect, cut to shot of Laurie spitting out coffee)

Going through the "getting a triangle on screen in Vulkan" ritual again and the latest version of the validation layers show a bug in how semaphores are used across multiple frames, which appears in both the vulkan-tutorial and apparently was in vkcube from 2014-2025. I'm not trying to bash either of these sources at all, but I do think there's an argument to be made that if the hello cube sample project from Khronos wasn't able to use Vulkan's synchronization model correctly and nobody noticed for 10 years, the API might be too complicated..

@Njord I put "useful in limited cases", but I'm aware that at least some of these cases are themselves caused by the rise of LLMs. For example as web search gets worse and worse, chat bots become relatively better in comparison, but the end result could still be worse than when we had working web search. Similarly the more AI generated garbage I'm presented with to read, the more need I have of a chat bot to summarise it for me.