The four organizations who maintain your favorite open-source DNS software, ISC, CZ.NIC, PowerDNS and NLnet Labs, gave a lighting talk at @dnsoarc 46 about the avalanche of LLM-assisted security reports for their projects, and the effect it has on us and our users.
The last slide ends on a “Hug your OSS maintainer" note, but I think this is understating the gravity of this situation. I hope we put forward a stronger message during the repeat of this presentation at RIPE 92.
People need to consider that we are in a situation where developers with talent, purpose and experience have created something valuable for the internet community over the last 20+ years. They could have chosen to work at $MEGACORP for twice, three times the pay, but they chose to do something meaningful.
Now, the body of work they carefully designed and maintained over the last decades is being picked apart by an LLM. Yes, as a result the products become some definition of “more secure” but there is no reasonable prospect that this avalanche of reports will end. Ignoring them is not an option. Feature development has come to a halt.
As an employer, what am I supposed to tell my developers? Thanks for creating this amazing DNS software over the last 20 years, it looks like you’ll spend the next couple of years triaging and fixing bugs and coordinating CVEs with your peers.
How do we keep people motivated to do open source and even if we do, how do we keep this development model sustainable? We can’t pivot to the ‘agentic era’ just like that and even if we could, I think my colleagues do this job to create something amazing—artisanal if you will—not to to maximize output at all costs so shareholders get rich.
Practically though, encouraging organizations to purchase a support contract will certainly help on the short term, because:
- You will get access to world class support;
- You will get early security vulnerability notices under NDA, keeping your critical infrastructure safe from a whole new class of LLM fueled risks; and
- In the grand scheme of things, you will help keep this open source model sustainable so your favorite DNS software continues to exist and thrive.
new Linux LPE, this time using io_uring. I think it's good to point out this Google security blog from 2023. Sending all kernel commits into an LLM might be somewhat new, the mitigation measures are already old.
also I think io_uring is a very cool technology, it just has it's tradeoffs like everything. That's all :)
English is a funny language… terms and phrases can mean more than one thing based on usage.
Take “I’m apolitical”, for example.
When it expresses that you only care about this quarter’s profits while the planet burns, it means “I’m stupid.”
When it’s that you can’t be bothered about the plight of people less fortunate than you, it means “I’m privileged” and “fuck you, I got mine.”
And when it shows that you’re neutral about – and thus actively complicit in – genocide, it means “I’m evil.”
What a language, eh?
it's great when you see a project you've been following for years has added a CLAUDE.md/AGENTS.md file to the repo, and you have a little moment of sadness
then you click it and it says "AI must not be used for this project" and goes on to explain why
Move fast and break things means you have to fix the broken things. Move slow and make things means you have cool stuff to play with for longer. I think the choice is obvious.
If this doesn’t get GitHub to take security seriously, nothing will: https://xcancel.com/github/status/2056884788179726685
RE: https://zeroes.ca/@VeeRat/116603185859484562
The issue I'm seeing is, that criticizing AI for the harm it does equals criticizing a good friend of some and their natural reaction are along the lines of "they are a good person who help me a lot! It can't be!" and it becomes impossible to talk about it, because many, for that reason alone, just don't want to.
And yeah, those AI companies totally exploit this on purpose, I have no doubts about them knowing exactly what they are doing there.
2 weeks ago I gave a talk about postmarketOS at the spring NLUUG (Netherlands Unix Users Group) conference.
It's in English so should be accessible to all!
https://ftp.nluug.nl/pub/video/nluug/2026-05-07-vj26/postmarketos-av1.mp4
RE: https://mastodon.social/@cmconseils/116589464367576074
But we need AI to make the investors happy!!! 📈
(maybe smth is wrong elsewhere that needs fixing...)
Heads up for edge users, OpenRC may hang on boot or shutdown with Plymouth.
"A fix was submitted to OpenRC, but in the meantime the Plymouth OpenRC integration plugin has been temporarily replaced by a kill-plymouth OpenRC service that runs on boot, just like what used to be the case with pbsplash. This workaround is included in postmarketos-bootsplash 0.5.1."
https://postmarketos.org/edge/2026/05/14/plymouth-openrc-issues/
Edit: The fixed package has been built and is available in the repos. It is now safe to update.
Heads up for edge users, a recent losetup update broke the initramfs.
"This was fixed in postmarketos-initramfs 3.10.2. If you upgraded your system recently and are not running that version, **do not reboot** until you upgrade to postmarketos-initramfs 3.10.2"
It is important to note that the upgraded package is not built yet. If you would like, you can follow https://build.postmarketos.org/#queued to see when it is ready.
https://postmarketos.org/edge/2026/05/15/losetup-option-removal-leading-to-boot-failure/
New stable kernels landed today in Alpine Linux. They address ssh-keysign-pwn (CVE-2026-46333)
- 6.18.31
- 6.12.89
- 6.6.139
- 6.1.173
- 5.15.207
These have been backported all the way down to Alpine Linux 3.16.
Edit: These versions fix ssh-keysign-pwn, not Fragnesia