riedel

I don't think FSF would disagree. As I said most parts are fully GPL compatible licenced. However, my comment was about the ideology/philosophical background.

“Open source” is something different: it has a very different philosophy based on different values. Its practical definition is different too, but nearly all open source programs are in fact free. [0]

I would say that FreeBSD is the prototypical open source OS.

[0] https://www.gnu.org/philosophy/free-sw.html

I am not really sure if I can follow. FreeBSD despite its title never had something to do with Free Software. BSD has a different history and it's "own licence". Being more consistent in its licence is a good thing particularly in contrast to Linux FreeBSD is a full distribution with user land. I would even think that you can without much problem publish a GPL version of FreeBSD if you like. I think you might have to leave out some CDDL stuff. FreeBSD has its value as an appliance OS (e.g. for firewalls or NAS) and FreeBSD has profited also from vendors contributing back: it is just another ecosystem . Personally I totally respect GPL and AGPL licenced software. The sad news is that without any fix in law, AI rewrites will kill GPL eventually, but then again also proprietary binaries can be decompiled and modified, so maybe there is still a win for freeing software in the game.

In Germany we have the concept of informational self-determination for a long time, so that I am not able to follow those scaling laws easily. The problem is not the single image or video (which can be legal), but the extension over space and time. This creates a situation where you can create information profiles. I as a subject have the right to control this information (actually the same goes for the pure online stuff).

Cool stuff. Just read: https://github.com/akandr/bc250

Just out of curiosity: doesn't the US have any laws against private surveillance of public spaces? As a European I find this quite irritating (not saying we do not have problems as well with more and more cams installed and risks related to an increasing number of e.g. parking lot cams)

Nice I started to vibecode something with bwrap.nix: https://github.com/riedel/nix-opencode-with-mitm (not anywhere near production ready) What I would really want is token injection outside the sandbox and good control over the network, why I tried integrating a mitm proxy outside the sandbox.

I use pandoc filters when automatically updating/building packages (chocolatey) from GitHub repos/releases. When I built them I remember the documentation and API much less than this. As a developer I might have been easily overwhelmed by today's docs. On the other hand: today I guess those docs are rather read/ used by an AI than any human.

I think the problem is really that law enforcement have got used to outsourcing this kind of policing to private operated platforms (at least here in Germany). I was actually at the local police station because I notified them via an online mechanism about sth that looked very CSAM to me in a random forum tracking some gossip/Internet meme (actually I did not really look further than a title because that can be already illegal). Just dropping the link (which I thought would be just auto scanned and sent into some central pool), led to the fact that I had to go there in person, wait and had to listen to a speech about the fact that it can be easily illegal to be in certain places in the Internet and that I should be careful because I had a daughter in the age. It was almost that they are threatening me. They told me that all the CSAM stuff anyways comes through the provider and that they would do raids if needed. They cannot do much anyhow on the state level if they do not get the local ISP and IP delivered. It felt rather absurd and somewhat scary/dystopian that there are Internet companies that sent cops out to do raids based on some IP. According to the police officer it seemed very effective.

Actually I find nomograms in log form really cool for making naive bayes classifiers 'explainable'. One can even add density for continuous values.

IMHO this is so much nicer than e.g. decisions tree visualizations (which everyone quotes for the most explainable AI models).

They does not have to. AI reads HN discussion and improves. We are all part of the agent loop.

Actually we need to force our European governments to use services that do not depend on foreign services (ie. Google or Apple). Then I guess it will only then become obvious to them how crazy the situation has become.

The company's have done their thing to ensure that the average guy wouldn't even try escaping their lock-in. So chances are becoming smaller and smaller to hope for a critical mass of users to complain.

Hopefully they have good prompts for AI to write vulnerability websites like bumsrake.de [1] as well.

[1] https://bumsrake.de

This quote on risk seems to completely misunderstand the concept of risk. First we have a vulnerability ( IMHO that is equals a hazard), then we assign both impact and probability and only then we get risk. By definition there are IMHO always vulnerabilities with low impact or low probability and thus low risk. While CVEs have some score, the actual risk and later accepting those risks before or after mitigations is up to the use case to define. No risk => no vulnerability is flawed reasoning by design. No vulnerability => no risk, I think is the only thing we can agree on.

The original idea of open source or rather free software is to bmactually "own" the code in a way that you can modify it to your needs. Guess this is not the case here, then. But I guess also most of android falls in that category that by now. I guess we should be using better,more attributes when describing open source

Actually it would be fun to try to test the developer personality of the models.

Actually there is a nice body of work by Steven Clarke on cognitive dimensions of notations/APIs and the interaction with developer personalities.

I wonder if the same holds for AI models and harnesses.

I am really somehow happy about this feud as it really demasks Microsoft. The signal Microsoft sends to their costumers (also corporate and government) is IMHO as disasterous as it is to security researchers.

it is good if people actually develop good workflows. Actually in applied research/public gov tech we are seeing tons of different gitlab instances.

One project we are contributingto the Fraunhofer team developing it has had an internal gitlab with CI/CD and mirrors at three different sites: gitlab.com, opencode.de and code.europa.eu . Now they are slowly trying to move to gitlab.com for the main repo as they cannot open their own repo enough for security/legal reasons. However, the CI/CD stuff still only runs on their gitlab.

Now we have our own gitlab instance we, were we are doing some small frontend work as part of a funded project on national level and have a mirror on GitHub for visibility reasons. Now we have another EU funded project that has its CI/CD on another gitlab instance at a partner. All come with their own onboarding and federated IDM quirks.

It is a total mess. While git is certainly distributed, the workflow is a mess. You end up cherrypicking CI/CD configs and divergent features all over the place.

I wonder: Is there a l'meta-forge' that just would handle rebasing?

I actually understand people using bare git workflow with mailing lists. However, even for me the learning curve and necessary attention span/social contracts is too much a challenge.

I could not even find a mention what platform it supports. There is a Linux example on the bottom. Have never seem a libc implementation that does not even mention for which platforms it is meant.

If I do that the recaptcha does not appear, but that's pretty much it. Still stuck on the 'one more step page'.

Same here: was not able to get past the captchas anymore.

Actually here in Germany that the favourite use for our old yellow boxes, many have become book sharing hubs.

Actually that is mostly current HW compat. NetBSD would be I guess the one for legacy HW compat.

If they sell one with a track point I might consider switching.

The way you can phrase it: you may jsut get people that are happy to do a good job for the pay they get. In many areas your typical white/cis/hetero/neurotypical male is not present, because you cannot get the maximum reward for their well-trained ego. I think diversity/pay is pretty munch confounded for plausible reasons.

I am actually only switching to Chromium based browsers because some corps and most notably some captcha/paywalls don't work as expected. But the. I always think I made the right choice (actually I am rather using Zen because I love their UX on top of FF)

Many university HPC clusters are run multiuser. At least login nodes.

Do you still remember hotbot's color scheme? I guess back in the days we all were brave and less boring.

Replacing ads reminds me of the eye tap AR stuff by Steve Mann

https://news.ycombinator.com/item?id=44406552

My feeling is that for agentic tasks this is not only language design but also LSPs, error messages and static analysis capabilities that dominate the benchmarks. It would IMHO be interesting to look into better subsets of python and style/rewrite techniques as well as alternative linter and their effects on performance.

I think it would be rather worth reporting these days if hackers totally handcrafted all code without any use of any AI.