Simon B

@foobarry
760 Followers
317 Following
255 Posts

I lead a team running a #Supercomputer at a University in the #UK.

#infosec #privacy #FOSS #retro #hpc #tech #dogs #academia #uk #keyboards #CasioWatches #MetalDetecting #linux #nobot and #ScaleModelling #photography

Long-time fedi user, although i have recently moved instance

Bloghttps://cyb.org.uk/about
Per-screen virtual desktops in kde plasma just got merged after a bazillion year wait. It had a big bounty on it - Hope the pledgers come good on their promises.
https://discuss.kde.org/t/bug-fix-per-screen-virtual-desktops/16241/83
#kde #plasma
[Bug Fix] Per-screen virtual desktops

There is no such think as a “KDE team”. There are KDE contributors (of which you are part of) and more seasoned/regular ones. Each time I see “KDE team”, I wonder where it comes from, it has been invented. Contrary to a team, there are no captain, no barrier to entry, “no side” (we only want to push forward FOSS according to our manifesto). I guess, just like people refer to KDE as a desktop which hasn’t been true for 10+years, KDE is a community; they hallucinated this, pun intented. KDE C...

KDE Discuss
Comparing the steps recorded by the skmei 2596 on my left wrist Vs the mi band 7 on my right wrist today. #skmei #watch

The results of the 26.04  Xubuntu Community Wallpaper Contest have been announced!

https://xubuntu.org/news/releases/26.04/2026-04-10-xubuntu-26-04-community-wallpaper-contest-winners/

Thank you, everyone, for your wonderful submissions and for voting for your favorites.

#Xubuntu #ResoluteRaccoon #WallpaperContest

when chewie met chewie #starwars
Surely I can't be the only person who follows the rule: "never trust a person who dresses like a magician"
vscode/cursor/claude are horrible software that are really not designed with networked filesystems in mind (or they don't care). apart from the obvious security implications of extensions running on the remote server, it is regularly scanning 20,000 files per second with ripgrep and fileWatcher. unfortunately users can't be relied upon to implement recommendations so I need to find a server side throttle to use.

When I start reading a book review I'm now getting quite good at detecting if these words will appear at the end:

"Thanks to netgalley and the publisher for sending me this advance reader copy for review."

Try asking an LLM not to speak until a trigger word is said. It will fail abjectly, and give all sorts of reasons why it doesn't need to follow the simple instruction it agreed to. there was nothing unsafe about my instruction, but some design rule overrode the instruction despite claiming that it would comply - in fact the bot suggested this method in the first place.
Extrapolate this issue to real world tasks and you have a big problem.

Twitter: opinions are mine, not my employer's

Linkedin: opinions are my employer's, not mine

Mastodon: opinions are from my fursona

I don't like that people can abuse the Freedom of Information request system with demands to populate their sales database. The effort on their part compared to the effort required on our side to respond is vast. #foi