My work is mostly just Claude Code and Codex, but always in a Docker Sandbox [1]. I find it almost the only reasonable way to run the harness safely, never run those on my laptop directly.
I am now working on reproducing a similar setup in a cloud VM + LXC, as opposed to running it on my laptop. The idea is to be able to spawn and provision disposable VMs with some automation and let them work in the background without a need to keep my laptop on `caffein[at]e`.
OpenHands [2] seems like an interesting solution in this setup, too, but I am not yet sold on it.
To be fair: while critical (9.8) it has an EPSS score of <1% (probability of it being actually used). According to [1], [2], and others, at worst the issue results in a kernel crash or a denial of service. I read the post by Anthropic too, but something just doesn't reconcile there between everyone saying its a DoS, and Anthropic alone claiming RCE. So far the advisories I read state "under specific conditions ... result in a kernel crash" and similar.
As the days pass the probability of it being actually used is increasing. The EPSS score is clearly just an estimate, and according to the nist cve database, attack complexity is low so I would take the EPSS score with a grain of salt.
Something like this is rowhammer. Can it be exploited in a lab setting vs practically in the wild? Is fail2ban + WAF sufficient? How much network traffic is needed to remotely trigger the exploit in the field? How much inside knowledge of the system itself needs to be known?
These go from extreme edge cases to viable approaches over time, so it needs to be patched. I think the really messed up thing is the lack of communication from the Ubuntu team. If they have a rationale either way, that's fine. They may disagree on the severity, identified a problem in the fix, etc, but the lack of transparency is causing a loss of confidence in the team's abilities.
When you change the language you are essentially moving to a different Wikipedia site, so yes it does - it pulls the results from the local site you are on.
Same for "AI". LLM is about the only kind of AI we talk about now, so the two blended. Nothing dramatic I guess, but it bothers me when tech people, who are supposed to know the difference, confuse the two.
There are more details on the official website [1] and in the official video [2]
I already love it. The infotainment is proudly named Radioën. They made the champignon pedals!, although I am not sure this will be kept in the production model due to regulation. The seats will definitely need headrests in the 21st :)
And there is a van model. I doubt that one will be a commercial success, though.
I could actually see the van being more successful than the car, the city delivery vehicle market has gotten pretty dire and the electric ranges are awful so it’s not like the 2cv lcv would have to compete much on that.
That link [1] is an absolutely awful page from a usability perspective. There's no other way to navigate the site than with the scroll wheel, and the amount of scrolling required makes my finger hurt before I've even reached the end of the first segment. It also has no anchors and does not remember reading position when refreshing or navigating back.
> My office was wired with two sets of electrical sockets, one connected to grid mains and the other connected to the UPS (for important equipment only).
This I think is a standard practice in any large office building in France (Europe? Global?) too, regardless of the frequency of the power cuts. I don't remember witnessing any power cuts in the office (not since long ago anyway), but in most of the places there are systematically white and red sockets, red ones being powered through a UPS. So, server machines go into red ones. Well, we don't have servers anymore, but the sockets are still there anyway. Hopefully UPS too, LOL.
No, this isn't remotely "typical." You have no idea what you're talking about.
The building might be on a generator. The whole building is most assuredly not on UPS.
Even in hospitals only certain outlets are on UPS - the stuff that absolutely cannot be interrupted - though a lot of hospital equipment now has its own battery system, at least in part for convenience.
I don't know if it's still done but semiconductor plants used to use intertia-based grid smoothing/UPS systems to handle millisecond-level interrupts and voltage fluctuations.
Agreed. Nothing actually typical at all for regular "just offices".
I do remember working ~25 years ago at a place that did have those two kinds of outlets. Our computers and monitors would be on the UPS+generator ones, while desk lamps, office lamps etc. were not. But it was specifically because the offices were on-site w/ the two data center complexes the company had and the building I was in happened to also be wired into that whole setup. Not every other one was either, only that one and the 24/7 NOC.
I only knew this was a thing because my office mate's computer was in the wrong kind of socket. One day his lamp stayed on while the computer turned off and for me the other way around. Soon after we heard the diesels starting and had to close the window coz of the fumes.
At full capacity of the data center, the absolutely huge UPSs could run the data center for all of 10 minutes. So that's how long the diesels had to actually start and take over.
I've never seen them in France. I've seen them in a chemistry lab section of a university in New Zealand and I believe they're also common in data centers, NOCs, etc. worldwide.
I think it depends on the building. A typical inner Paris office (converted Haussmann building) won't have it usually, but traditional larger offices I have been to have sockets like this one for example: https://www.legrand.fr/pro/catalogue/triple-prises-de-couran...
> It's almost like grid inertia is a different problem from matching changes in demand, and nuclear really sucks at the latter...
For the recent big outage in France, neither of those was the problem. Some big transformers overloaded and there wasn't enough redundancy.
> Now everyone has all their eggs in the Amazon/Microsoft/Google basket, and instead of the (ahem, regulated) telco or electric company going out...the cloud provider does, and unlike before where you could still work locally - now you can't do anything.
> But hey, it was cheaper, right?
> Oops...
I'm not sure what you mean by "oops". Those datacenters do have downtime but companies seem to be happy sticking with them anyway. It seems like it still saves money.
Does anyone have a real-world usage feedback on this model? Because the benchmarks look, I would say, too good: it scores even higher than Qwen 3.8 27B in some benchmarks - I am not even sure how is it possible. So the question is if this is benchmaxxing or is it actually good for practical usage like coding in complex projects, research, etc.
But I DO record some games so that I can replay them later. That way if you log on and you're the only person, you can still watch the ghost of a past game. This isn't relevant right now (there are 500ish people playing! wow!)- I only show past games if there aren't active ones to show.
Hello! Great job on the project! After seeing your blog, I actually decided to go to that 30th Anniversary after party! I didn't even know about Rhizome. They are such a cool group!
Man I was NOT dressed as well as everyone else there and had no idea what I was doing but it was still fun ha ha! In hindsight, I guess it was everyone coming back from that benefit dinner?
Was your exhibit supposed to be there? I didn't see it on either floor.
I am now working on reproducing a similar setup in a cloud VM + LXC, as opposed to running it on my laptop. The idea is to be able to spawn and provision disposable VMs with some automation and let them work in the background without a need to keep my laptop on `caffein[at]e`.
OpenHands [2] seems like an interesting solution in this setup, too, but I am not yet sold on it.
[1] https://www.docker.com/products/docker-sandboxes/ [2] https://www.openhands.dev/
reply