I liked Multipass until I ran into storage corruption issues like many others [1][2] and broke all my VMs. On MacOS, Multipass uses qemu in which the apple-hvf support isn't getting much updates after 2021 [3].

[1]: https://github.com/canonical/multipass/issues/3455 [2]: https://github.com/canonical/multipass/issues/3235 [3]: https://github.com/qemu/qemu/commits/master/accel/hvf

The VM disk corruption problem has been affecting all open source Apple hvf hypervisors. lima with vfkit or UTM are most promising ones for now: https://github.com/lima-vm/lima/issues/1957#issuecomment-182..., https://github.com/lima-vm/lima/pull/2026, https://github.com/utmapp/UTM/pull/5919

Thanks for the heads up.

Re: quemu and apple hvf that’s really sad news.

Somewhat related talk: Modernizing Compiler Design for Carbon Toolchain - CppNow 2023 (https://www.youtube.com/watch?v=ZI198eFghJk)

This talk was enormously educational when I first dove into LLVM and modern compilers last January—Chandler is a great speaker.

(National) Security theatre https://en.m.wikipedia.org/wiki/Security_theater

Plus everyone who's involved will get richer.

The backdoor (test binary blob and autoconf) is not part of the pull request.

There is zero web presence for this person and associated email address.

Looks more likely a fake identity than compromised account.

Actually the "jiat0218" user part in his email address jiat0218@gmail.com has a bunch of matches on Taiwanese sites:

https://char.tw/blog/post/24397301

https://forum.babyhome.com.tw/topic/167439

https://bmwcct.com.tw/forums/thread1828.html

I think it's just a coincidence.

- All the posts are from 2004/2006. - "jiat" can be abbreviation for many common Chinese names.

I agree, probably a coincidence. Just wanted to point out we can, actually, find the username online.

It might just be a coincidence, but the same username from that gmail account also appears to have a Proton Mail address

I think it's not a coincidence: Hans Jansen (hansjansen162@outlook.com) has a matching account on Proton mail too (hansjansen162@proton.me). Furthermore, the Outlook account is configured as recovery e-mail for the Proton account.

This is all I can find on them.

    carrd.co jiat0218@gmail.com business https://jiat0218@gmail.com.carrd.co
    eBay JiaT75 shopping https://www.ebay.com/usr/JiaT75
    giters jiat0218 coding https://giters.com/jiat0218
    giters JiaT75 coding https://giters.com/JiaT75
    GitHub jiat0218 coding https://github.com/jiat0218
    GitHub JiaT75 coding https://github.com/JiaT75
    Mastodon-meow.so.. jiat0218@gmail.com social https://meow.social/@jiat0218@gmail.com

Beyond that, nothing surefire. (This is all publicly queryable information, if anyone is curious).

JiaT75 also used "jiatan" on Libera.Chat using a Singapore IP address (possibly a proxy/VPN).

Where did you gather this information from?

I've never had a web presencse for my associated emails due to wanting to avoid spammers. I don't have a false identity.

Keep in mind that having a "false identity" does not make you a malicious actor. I have a serious project I work on under another pseudonym, but it has to do more with the fact that I do not want my real name to be associated with that project AND having a serious case of impostor syndrome. :/

That, and I used to contribute to various games (forks of ioquake3) when I was a teen and I wanted to keep my real name private.

Someone named "John is good" claims they aren't a malicious actor... You're trying real hard to convince us, huh.

Oh yeah, I am using a pseudonym here as well, because I have controversial views in some topics. :P

> I don't have a false identity.

That's just what someone with a false identity would say.. get him boys!

The biggest /S

I am more interest about his git commits https://github.com/JiaT75?tab=overview&from=2021-12-01&to=20... If JiaT75 is a Chinese, then his working log should follow Chinese Holiday, especially Spring Festival and National Holiday. Chinese usually not work on first 3 days of Spring Festival and National Holiday - 2021 2/11 - 2/13 (few commits), 2021 10/1 - 10/3 (nothing) 2022 1/31 - 2/2 (huge commits on 1/31, suspect), 2022 10/1 - 10/3 (nothing) 2023 and 2024, not very much commits. So 2022 1/31 huge commits is a proof that he is not follow Chinese holiday.

But wait, 2021 is his active year, but he missed almost all Aug. Is he on holiday? Who can have such a long holiday? What i can think is a solider who has a long vacation (探亲假). So let's guess he is a solider then it's sense that he worked on Spring Holiday because they need on duty. Let's double check again, if he is a solider, then they will have a holiday on every Aug. 1 because it's liberation army day. I check and no commits on all 4 years Aug. 1.

Did you check Chinese social media?

I found this link on Zhihu: https://www.zhihu.com/question/650826484

Why would you think the person would have social media (or would even be on Chinese social media specifically), given the sophistication and planning?

I mention Chinese social media specifically because I know it's not indexed so well by western search engines. You can't conclude someone has no social footprint until you've actually checked.

Regardless of how likely you think it is, finding a social media footprint would be useful information. Seek information first, reach conclusions second.

i wonder if that avatar, familiarity with C/C++ and Git, and "offering help with open source projects" is just coincidence

https://github.com/JiaT75

https://twitter.com/JiaTan1337/status/1774931375994319244

kind of interesting also to see this account was set up ~2 months ago. if it's a troll, it's a somewhat poor joke.

Political theatre? Courts will most likely strike it down for violating the First Amendment.

That could prove to be an empty hope, considering the fact that kids already don't have second amendment rights. There is also a long precedent of allowing bans on the medium (spraypaint, public parks and megaphones) that do not mention anything about the specific message.

Really interesting!

Curious, why reimplementing jax in rust (nox)?

For me, it boils down to 3 reasons. One is that the JAX tracing process can get a little slow for more complex programs. By bringing Jaxprs into Rust, we can optimize the compile path more. This doesn't matter for small simple programs, but for larger programs (especially with lots of unrolling) it could matter a lot. The 2nd reason is that I just prefer Rust for making production-ready software. The strong type-system especially allows you to catch many errors at compile time, Python is moving in that direction, but IMO isn't there yet. The 3rd reason is that we want this software to easily integrate with your existing flight software (the control software for your drone/satellite), and using a "systems" language like Rust makes that easier. Eventually, we would love to build out a suite of flight software in Rust that can easily integrate with the simulation but is still flight-ready.

A fun side-effect of this architecture is that when we run your simulation we aren't actually running your code, Python or Rust in the loop. We compile down the whole thing to HLO at runtime, then run that HLO. So we could theoretically support lots of other languages.

Exciting! I have been wanting a Rust port of JAX for a long time! (I mostly care about the ability to write numerical GPU code via XLA, not differentiation)

Is there a crate in the works? I would start using it (and contributing if needed) instantly.

Yes you can use Nox separately today. https://github.com/elodin-sys/elodin/tree/main/libs/nox . It just hasn't been released on crates.io yet, and no promises about stability. I mentioned this in a sibling comment, but we likely will have to rename Nox to something else since someone is squatting the name on crates.io

I'd love to see someone who has the resources train a model bigger than 2.8b and show the scaling law still holds.

Some prior comments said those architectures lack the memory or something of a transformer. That there’s a weakness that’s keeping people using transformers. If true, I’d like to also see tests of various domains with equivalent transformer and Mamba designs to see if that difference impacted anything. From there, we’d have a better idea about whether Mamba-176B is worth the money.

Language Modeling Is Compression: https://arxiv.org/abs/2309.10668

From https://arxiv.org/pdf/2402.08268.pdf

> We trained our models using TPUv4-1024, which is approximately equivalent to 450 A100s

> Inference for such long sequences requires a minimum of v4-128

So you'll need ~60 A100 for inference.