It's interesting hw lng it took for linux to get a user friendly application firewall like OpenSnitch

It's because there's no way to make universal kernel modules/drivers, like it is on Windows.

The way to make kernel modules is to submit them to the kernel. Not really sure what a “universal kernel module” really is.

Also that seems irrelevant because it seems this was implemented in eBPF so no kernel modules are required.

> The way to make kernel modules is to submit them to the kernel.

Then it would need to be published as GPL, but with no guarantee that it will ever be accepted.

> Not really sure what a “universal kernel module” really is.

A .ko that can be loaded in a wide range of kernel versions.

> Also that seems irrelevant because it seems this was implemented in eBPF so no kernel modules are required.

And it has serious limitations. There's a chapter of them in the readme.

I see both sides here. Wanting to preserve your natural voice is valid, but editing and using tools don't necessarily take that away. In fact, they can help make your intended message clearer. It probably comes down to how much control you keep over the final result rather than wheater you use tools at all.

What annoys me here is that people say "I use AI as style checker to make my writing better" or claim that good writing is unfairly judged as being by AI ... and then proceed to describe inferior writing results they achieved with AI. None of what author wrote there signals that the way he uses AI made his writing better. His use of AI made his output inferior. And not just in a the "loosing own voice" way worst, but literally in the "the final text is less effective writing".

I do not mean this comment to be kick against AI. It is very good for some stuff, it is less good for other stuff. What annoys me is someone calling output superior while actually complaining about it being inferior.

Hey, maybe that llm needs to be used differently to achieve actually good writing results.

yes sir