LMAO, can we just issue one giant CVE for the entire Linux kernel

PoC for CUPS LPE on Linux https://github.com/v12-security/pocs/tree/main/cups/cups2root

@buherator i remember reading during the second iraq war, operations led by the british had far fewer friendly fire casualties because they had not just a system for putting a case for a target together, but also researchers trying to rule out the validity of a target (that's a friendly, that's a school, etc)

the americans didn't have that

...then they automated it

and i guess, *now* the stakes are high enough to give them a target offramp, huh

@frobozz My prediction is that this will be covered by some bullshit about "adjustment" then a few million will be spent on additional guArDrAILs then the next hallucination will cause actual damage. So yeah, not a bad bet.

The future will be much more embarassing for humanity than Skynet

US military had close call after using AI for false intelligence report

https://edition.cnn.com/2026/09/18/politics/us-military-ai-false-intelligence-china-ship

They almost attacked a Chinese ship believing it was transporting nuclear weapons component (/via OSINTTechnical@X)

Sweden is also choosing @forgejo for their digitally sovereign government codeplatform, just like the Dutch code.overheid.nl 🥳🥳

https://github.com/diggsweden/ena-kodsamverkansplattform-poc

This is a nice step towards less dependence on US big tech, aka Github.

In my last days at the @BZKopensource I hope to set up a working group with other EU governments, to work towards a federated network of codeplatforms 🇪🇺

#codeplatform #forgejo #ospo #digitalsovereignty #opensource #foss #govtech

Buyers of pervert glasses join the war on pervert glasses: https://www.latimes.com/business/story/2026-09-18/meta-glasses-captured-shared-intimate-images-without-users-consent-lawsuit-claims

The quickest teardown ever. Zero screws removed. 👀 Your first look inside the iPhone 18 Pro thanks to Lumafield!

just in case you don't have it saved, here's the most important reaction meme for navigating the fediverse

Ramdom thought / further evidence that this universe is against me:

In the cycling world, clipless pedals are pedals that you clip into.

I don't like this.

@freddy I was hoping this is the Hungarian band Linx, still perfect for Fri thx:) Do you know Wizo's split with Auróra?

https://www.youtube.com/watch?v=_0xcwW3CqIQ

🎶🎶🎶
Was ich heute kann besorgen,
verschieb' ich heute schon auf morgen.
Doch meistens lass ich's ganz oder bekomme es nicht hin.
Auch wenn ich darauf wette,
krieg ich's nicht auf die Kette
🎶🎶🎶

Bitte einmal anhören, alle. Zeitmanagement von SOKO LiNX & WIZO.
https://www.youtube.com/watch?v=OJgY-TRuW7s

@freddy Oh yeah and this other thing called the clipboard...or the keyboard...

@freddy sorry what? :)

Posting to social media is like writing to the wall of a public toilet.

If you are bothered by replies, pictures taken of your work, or genitals drawn next to it, then you are probably using the wrong medium.

I've turned off pull requests from non-members for Flask, Werkzeug, Jinja, and Click this week, and have enjoyed not waking up to a fresh wave of AI junk every morning. I think I'm going to update our contributing guide to say "introduce yourself and how you use the library in chat first if you want to contribute pull requests". Sad that we had to go here, but aside from conference sprints we weren't really getting outside contribution anyway (which is also sad, but that's been the case a lot longer than AI has been around). #python #flask #opensource

@tychotithonus I miss OS diversity too, guess it'll be a future improvement? Prioritizing Linux instructions makes sense if you consider CI pipelines.

My friends, I have spent months assessing "AI security" products, especially ones that claim to prevent jailbreaking, prompt injection, etc. Here's what I know:

1. They don't.
2. These companies are reinventing EDR from first principles, but with minimal understanding of system internals.

It's really a sight to behold.

@HalvarFlake Thanks for highlighting the importance of statistical rigor! I'm exactly as bad at stats as you describe, but seeing conclusions drawn from 3-5 test runs being the "standard" for model/prompt/agent evaluation feels just wrong.