Congrats gaeld and team

The demo is very impressive!

disclaimer: I've known the founder for a while, as legitimate as it gets in deep tech, real years of research and engineering behind this, not vaporware

> For monitoring I use and recommend UpDown.io, which doesn’t seem to be listed there.

It doesn't seem very well known, but I've been a happy user. Most of the others have become over-bloated with a shitty UI.

> Nicholas Carlini, a researcher at Anthropic, orchestrated 16 parallel Claude agents to write a production C compiler in Rust.

To write a proof-of-concept C compiler, not a production-grade one...

Hard to take the article seriously after this

To be fair it was a totally unattended zero shot loop developed compiler - which is pretty remarkable no matter how you cut it.

I’m surprised what made you quit reading wasn’t the Claude voice sneaking through their half success attempt voice clone.

Yeah and

> A C compiler written in Rust used to be a graduate thesis. It isn’t anymore.

Or maybe like a little recreational project for multiple weekends.

There is that weird myth that writing compilers is super hard. Writing a toy C compiler is not that big of a deal. It is a pretty simple language.

Now production-grade is another beast but that is something AI can't do.

You misinterpreted the comment you are citing.

This non-determinism would not and did not cause replays to diverge (the PRNG seed was most likely stored and would reproduce exactly the same results).

Your AI powered comment is wrong. Le monde has been doing this for years. They have a series of articles about this. There is no "gap closing."

it cannot email your secret key to an attacker because of prompt injection etc.

Almost a month old, original source: https://cybernews.com/security/global-data-leak-exposes-bill...

and I've never seen any confirmation elsewhere

Looks like CyberNews have edited the article with more info since first I saw it, it used to look quite suspicious and untrustworthy, it now has more info. Still doesn't say exactly what a record is, or how many uniques there are.

I presume the database exists, but some of the details don't add up. IDMerit say "IDMERIT’s systems and security infrastructure have never been compromised", "there has never been a data breach or exfiltration from [our partners'] systems during, before, or after this event" and "IDMerit does not own, control or store customer data". But Cybernews says that they "promptly secured the database" after being notified. Cybernews also didn't give the reason why they thought this was to do with IDMerit (unless I missed it). I can't quite make head nor tail of it.

It's a weird article. For one, the researcher says "they believe" the data belongs to IDMerit but apparently aren't sure. IDMerit denies it's the owner of the data nor is it any of their partners. And there's very few details about where or how they found this database. It's possibly some kind of hoax or ransom attempt? Or there's really just billions of unaccounted databases of private data just sitting all over the Internet.

The cybernews article does have some screenshots showing names like “idmb2c” … also that IDMerit was contacted in November and the ports were closed a day later.

To sum up the updates in the article

  - IDMerit asked the security researcher for proof, the researcher asked for money first, so IDMerit balked
  - IDMerit basically says they have no proof they were hacked, so they weren't
  - The researcher is a freelancer... for CyberNews...

Even if somebody followed up with IDMerit, it's likely they will say they are not affected. The security researcher is probably the only person who could prove whether they were or not vulnerable, at this point. If they don't come forward, we can only assume they weren't vulnerable, but we don't know. This is a good lesson for responsible disclosure in the future.

...also, this is yet another example of why we need a regulated Software Building Code, with penalties for not conforming to it. If somebody is found to be hosting a public Mongo instance with no authentication, it should be reported to a state or federal agency, so that real penalties can be applied, the way they are for other code violations. And they shouldn't have been allowed to launch with that in the first place. It shouldn't be up to random "security researchers" to police businesses.

You're getting downvoted because you didn't read the article.

It is specifically about cleaning up the data by removing these 3 and showing a clearer picture of acceleration without these 3 factors.

Great series of articles!

Don't they?

https://git.ffmpeg.org/gitweb/ffmpeg.git?a=search&h=HEAD&st=...