The_Decryptor

Fable is out there finding bugs in these things

Ehh, that’s debatable. An LLM will gladly spit out a vulnerability report, but as LLMs lack any understanding or reasoning capability, there’s literally nothing guaranteeing that it has any semblance in fact or reality. They are just random text generators after all.

Going by Anthropic’s own numbers for things like Mythos, it “found” 29,439 vulnerabilities, when a human actually checked if the reports made any sense (ignoring validity) that dropped to around 6,200 reports, when they were tested for validity that further dropped to around 1,300 reports, and when it came time to actually patch the issues only 516 made the cut (i.e. the rest might have been issues, but weren’t considered exploitable, etc.).

516 out of 29,439 is a 1.75% success rate, not exactly groundbreaking.

Now to be fair, according to Greg KH at a recent talk, Mythos is seemingly improving. Of 79 vulnerabilities it reported, a solid 4 were real. That’s a 5% success rate, though it performed much better when it simply reported other developers bug fixes, something it did 11 times.

He gets paid more than enough to afford high end hardware.

The sleep stuff is part of the ACPI standard, it’s all documented.

It’s up to the motherboard/SoC to support it properly, and Microsoft has an “entire system view” of how Windows works and can integrate it deeply. Compare that to Linux, the companies involved mostly care about servers, and the distros are assembled by disparate groups that don’t have the hardware testing capability that MS or OEMs have.

It has been wild to see Linux users hope that the GPL is rendered unenforceable.

You got a spare $80 million laying around to train one?

Corporate stuff and things like IoT devices were already exempt from the law, they didn’t need any changes.

iiuc it supported a nicer workflow than git did, but that’s not out of the ordinary as so does Darcs, Pijul, Jujutsu, Mercurial, etc. git is just “good enough” for people to put up with the rough edges, rather than learn a new tool that never has them to begin with.

It just never had the dev mindshare to catch on, it’s been a GNU project for over 18 years at this point and that hasn’t helped.

More individual participation, less corporate involvement, zero corporate ownership of essentials.

I’ve got some bad news about Linux then, the main contibutors are companies, and the infrastructure is owned by them.

It’s not individuals who paying Linus $1.5 million a year (Under “Compensation”).

Using an LLM to review patches does that, which is why he was receptive to the Sashiko bot.

Funnily enough, I was just reading a post from a dev complaining about the review bot and how it makes developing patches harder.

Previously review comments were from somebody who had experience with the code or subsystem in question, who understood how the whole thing worked. The LLM bot has none of that and will spit out nonsensical comments that the developer now needs to try to understand and verify, so they still have to ping other developers to see if the LLM output makes sense.

A box shaped computer? so pedestrian, mine is a hyperbolic paraboloid.

Defeatism just benefits them. We don’t need it.

It’s not defeatism btw, it’s opposition to the idea that people should have a say if their content is used to train an LLM.

Does systemd print much once it’s fully running? Checking my log the only chatty thing is networkmanager doing wifi scanning, 4 separate state change log messages per scan.

Yeah any decision would be on a case by case basis, which is normally something you’d want to avoid.

I’ve seen a couple of Linux devs talk about how they just give a prompt to claude and walk away leaving it alone to spit out the code, none of which can be licensed as GPL. But good luck working out what specific lines of what specific patches of theirs used an LLM vs. were re-written or such.

According to the dev, the biggest challenge was that the Motorola 68000 lacks a Memory Management Unit. So, a standard Linux setup with virtual memory would not work. Luckily, Linux still supports m68k systems without an MMU by using the uClinux/NOMMU option and a flat memory model.

I was honestly expecting it to be weirder hardware, turns out it’s a configuration the kernel already supports. Still neat, but somehow a bit of a letdown.

The DSP stuff is cool though, that’s sufficiently weird for me.

I’d call that plain engineering; you’ve separated out the independent logic from the point where it actually touches the system. That’s how stuff should be built and operate.

Try Ctrl+Shift+u followed by the hex value and then enter (e.g. 263A for ☺)

Reminds me of a local pro-privacy group I came across a while back, whose only public online presence was a Facebook group.

Like I get it from a “reach” standpoint, and I think it’s fine if you’re following a POSSE system, but I feel a lot of people don’t do that.

Are you talking about firmware?

Who are these people who think drivers can’t be open source?

I think it’s mostly maintenance issues, the maintainers being employees of the company that originally wrote and donated the driver, so there’s been a backlog of bugs that have sat unfixed.

This new driver on the other hand builds off the preexisting kernel support for NTFS, and the maintainer seems better equipped to respond to bugs in a timely manner.