GenAI must be regulated down but isn’t and won’t be any time soon. The matter of persisting existence of rogue AI is foreseeable at this point. This is the first big chance for Qubes OS.
Omarchy Linux has raised US$ 18.5 million in a short time, showing that it is possible for libre software to get the money they need if they position themselves accordingly.
I therefore suggest the following for the project (which won’t be done, but I believe in free will):
alphanerd(aka mental outlaw) devoted 1/2 hour to looking at omarchy, which I would summarize as an unreasonably insecure OS. If Qubes followed their funding approach I’d quickly be jumping to the paranoid about all things computing related camp. However, if a vision for reasonably secure agentic ai brought in donations (without serious strings attached) then I can see how omarchy could provide a model for what not to do.
Just for fun: I asked an AI system (Gemini Notebook) if the recent Berlin hack would have succeeded if they had used Qubes instead of Windows. It not only said this hack wouldn’t have worked with Qubes, but it even created a migration concept from Windows to Qubes, something the Berlin administration obviously can’t figure out on its own. So one might ask if this AI system is not more intelligent than some government officials.
The point of libre software is to hold on to the libre principles, and not to position themselves any other way.
More importantly, setting Omarchy as an example like that isn’t good IMO. Sure they raise money but I don’t like both the kind of money they raise, and the kind of product they have in order to raise them. KDE is a much better example, but even that I’m saying simply because I tried it a while ago and it seemed okay: I don’t track what developments happened afterwards, or happen now, or what developments followed funding.
TL;DR it may be possible for libre software to raise good money, but it doesn’t mean it should.
What do you mean by this? What adw says in the link looks very agreeable: no strict opposition, and yet no blind trust or favoring slop.
Enigmatic leaders have an awful tendency to only be good for publicity. As much as I like Joanna, and as much as I disagree with Marek, enigmatic leadership for its own sake is a bad idea. Also despite my Marek disagreements I must admit he is a smart guy and has good points. Separate person (not a leader) for public interaction could be better, but that is a different topic.
I haven’t seen anybody ever complaining about project name being similar to anything, nor specifically comic sans. What’s up with that?
I don’t keep a book of grudges. There might be one or two in the matrix chat or elsewhere but I keep them to myself unless I have something to show or the problem is objective.
In general there are plenty of bad things about QubesOS, I don’t remember which ones are Marek-induced, and I don’t want to search. Besides, I don’t want you to misinterpret what I really think - Marek tends to be the voice of reason. I agree with him much more than I disagree.
Well, without facts there is nothing to (mis)interpret. I asked just because you mentioned it twice in a single post, so I thought you may actually want to share what you are talking about in a less enigmatic way
Marek tends to be the voice of reason.
So does Joanna, although I have never had the chance to communicate with her. Along these lines, I don’t quite understand the notion of “enigmatic leader”, as I have not noticed anything that sounds like a riddle in her writings/speeches.
As a whole, the current thread is quite confusing.
I would like to sincerely apologize for the secessionist vibe that I mixed into the actual topic of this thread, which is how to improve the funding situation of the Qubes OS project by leveraging the current state of AI development.
I distance myself from my remarks that imply direct criticism of the leadership of the project.
As I am not part of any of the relevant networks, I know very little about how to benefit from this situation in specific.
But I do know that AI companies would like to solve this issue with, well, simple application of their own LLMs but due to the guardrails built into the freely accessible top models, they’re not really well suited for the task for they cannot easily distinguish between attacker and defender. I can’t imagine the situation becoming better with strong open-source models without guardrails either because they worsen the situation at the same time.
There gotta be some money in this for the project to get hold of right now or very soon.
