https://www.aftonbladet.se/nyheter/a/ArvG0E/cgi-sverige-uppg...

No. CGI has nothing to do with BankID.

IMO the most credible reports suggest that the source code and data involved are related to these four services:

https://www.cgi.com/se/sv/business-process-services/e-tjanst... "Mina engagemang offers a user-friendly and flexible solution that allows your customers to manage their cases directly through a personal portal. Here, users can view, track, and interact with their ongoing cases, which enhances both transparency and efficiency in the communication process." -- some kind of ticket/case management system for gov't agencies

https://www.cgi.com/se/sv/business-process-services/elektron... "With our secure end-to-end e-ID and eSign services, we can help you streamline document and contract management, gain access to all desired e-ID issuers, and improve cost efficiency." -- this sounds like a bad thing to compromise, but is to the best of my understanding a system for digital signatures on documents, and has no relation to BankID

https://www.cgi.com/se/sv/business-process-services/e-tjanst... "Gain better control over your organization’s representatives with our easy-to-use representative registry. By automating the identification and verification of representatives, you’ll gain a clear overview and enhance the security of your processes." -- sounds like some bullshit CRUD app for managing who can "represent" a gov't agency

https://www.cgi.com/se/sv/business-process-services/e-tjanst... "SHS is Sweden’s common standard for information exchange, enabling secure and efficient communication between government agencies, businesses, and organizations." -- this might be bad if real data was leaked

These are services used by various Swedish government agencies and it's pretty bad to have even a test instance of them hacked, but let's calm down. The entire Swedish state has not been compromised here.

That's incorrect. Skatteverket used CGI for BankID-login, I don't know if they still do. I have personal experience working on a BankID-login using CGI for another company and it is still active.

Edit: I just confirmed Skatteverket still uses CGI for BankID-auth. "funktionstjanster" is CGI.

Interesting how LLMs have their own preferences too. Those in particular are very often used by ChatGPT, while Claude until recently couldn't stop saying "You're absolutely right!"

I also have a problem now with "it's worth noting", I use it a lot, I still like it, but now it's a dangerous phrase because of LLM associations.

I have been consistently skeptical of LLM coding but the latest batch of models seems to have crossed some threshold. Just like everyone, I've been reading lots of news about LLMs. A week ago I decided to give Claude a serious try - use it as the main tool for my current work, with a thought out context file, planning etc. The results are impressive, it took about four hours to do a non-trivial refactor I had wanted but would have needed a few days to complete myself. A simpler feature where I'd need an hour of mostly mechanical work got completed in ten minutes by Claude.

But, I was keeping a close eye on Claude's plan and gradual changes. On several occasions I corrected the model because it was going to do something too complicated, or neglected a corner case that might occur, or other such issues that need actual technical skill to spot.

Sure, now a PM whose only skills are PowerPoint and office politics can create a product demo, change the output formatting in a real program and so on. But the PM has no technical understanding and can't even prompt well, let alone guide the LLM as it makes a wrong choice.

Technical experts should be in as much demand as ever, once the delirious "nobody will need to touch code ever again gives way to a realistic understanding that LLMs, like every other tool, work much better in expert hands. The bigger question to me is how new experts are going to appear. If nobody's hiring junior devs because LLMs can do junior work faster and cheaper, how is anyone going to become an expert?

It’s refreshing to hear I’m not the only one who feels this way. I went from using almost none of my copilot quota to burning through half of it in 3 days after switching to sonnet 4.6. I’m about to have to start lobbying for more tokens or buy my own subscription because it’s just that much more useful now.

I'm still not ready to sing praises about how awesome LLMs are, but after two years of incremental improvements since the first ChatGPT release, I feel these late-2025 models are the first substantial qualitative improvement.

Some people appear to have a capacity for learning, retention and understanding that is well outside the normal range. People like Ramanujan or von Neumann, or Tao. They learn at a speed that far exceeds the speed of what we would consider gifted students, they reach a deep and intuitive understanding of the material, and go on to make many discoveries / inventions of which even one would be enough for an ordinary scientist to be considered successful.

It seems there is something very different about their minds, but just what is it that allows those minds to operate at such a level?

They're a fantastic hardware company. But my admittedly very limited experience with Apple software, from iPad to their streaming service website, has been miserable. The UX doesn't work for me, the software just doesn't do what I want. Understandable, Apple very much designs their software to work for a particular workflow they come up with, if you like that workflow it's great, for someone like me it's miserable. But I would gladly buy their hardware if I could freely run an OS of my own choosing.

The assumption that this is a triangulated and well researched strategy doesn't match my experience in "real-job" world. I mean, maybe Apple is different because of their history, but I am not convinced anyone listens to anyone that articulates any math ideas beyond Algebra outside of some niche specialties because they don't understand it. And it's not that I'm some math god - I mean, that's what I studied, but there are people SO much more knowledgeable and capable, and they seem to get ignored too.

Like, I'm sure the guy who runs an insurance company listens to the actuaries about relative risk, but mostly, what I've just seen is someone makes a decision, and then finds post hoc ergo proctor hoc rationales for why this was a good decision down the line when they have to account for their choices.

For instance, it took my like a year at my old job, but I finally got most of the KPIs we were using to set strategy cancelled. The data we were using to generate those KPIs? Well in a few cases, after you seasonally differenced the data was no different than white noise. No autocorrelation whatsoever. In ALL the cases the autocorrelation was weak and it was all evaporated after a month or 2. You could MAYBE fit an MA model to it, but that seemed dodgy to me. And like, I'm not a major expert - I took 1 time series class in gradschool, and frankly, time series is kind of hard. But management had ZERO idea of what I was talking about when I was like, "hey, I don't think these numbers actually mean anything at all? Did anyone run an ACF?"

Then each month someone higher up the chain would say, "why is this number low?" And then they go out and search through the reams of data they had to come up with an answer that plausibly explained things. Was the number particularly "low?" No, it was within expected statistical noise thresholds, you are probably going to have at least have 1 number out of whack every 20 cycles or so... You still had to spend an hour in a meeting coming up with reasons for why it was low that went beyond "ummm, well, this is kind of random, and we'd expect to see this sort of thing ever couple years once or twice, we won't know if it's a trend for a few more months."

Anyway, this is a long anecdote to explain why I have no confidence that most companies do any sort of actual introspection. CEO creates targets and underlings build models that show how they're meeting or not meeting those targets. Now, hilariously, with Apple in particular I might be wrong, because in Tim Cook's defense, I'm pretty sure his education is in Industrial Engineering? So if any CEO is thinking about that stuff, it's him. Still, I am totally and completely unimpressed with the C-Suite sort of thinkers.

They're not dumb - like I've never really had a straight up dumbass manager outside of shitty lower jobs or small-mom-and-pop businesses? But I have seldom met any company that actually cared about the numbers - most say they do, but most just use those numbers to justify decisions they've already made.

Am I just unlucky? I'm I the witch in church here?

The environment is why I quit my job and started working for myself in January. I hated it. And not to sound like an arrogant ass because there were a LOT of way smarter people than me at $PREVIOUS_EMPLOYER, but having to have meetings to set our meetings, having to explain things that aren't statistically meaningful to people who don't understand stats anyway, and getting code reviews (when I could get them scheduled) from dudes who hadn't touched a keyboard in 5 years was... soul sucking? I'm not doing that anymore. Or ever again.

I mean, maybe it's because I had a more hands-on blue-collar adjacent job before I got into tech? Maybe it's because I'm a fool and couldn't play the game of "pretend to work and look busy. But - and I know this might be kind of messed up - I really like not having to explain things in a series of emails to people other than the customers. I really like not having to answer to anyone but my self and my customers. If I want to do something, well, I just do it now? That's a nice place to be. Riskier for sure, but I think the prior environment would have killed me, so maybe not.

Also, I have time to do shit that's interesting? Who would have guessed how much more time I'd have in the day when I didn't have 4.5 hours of meetings per day? Hell, I'm taking 2 classes at the university for fun (weird right?!) - I never could have done that before because I would have had to make a slide deck for Thursdays All-Hands or whatever and couldn't have missed the SUPER IMPORTANT MEETING that Jake has on the schedule that he'll show up for unprepared or just not show up to.

Nah, the hell with that. I'm never going back.

In the ancient days I used TLDP to learn about Linux stuff. Arch wiki is now the best doc. The actual shipped documentation on most Linux stuff is usually terrible.

GNU coreutils have man pages that are correct and list all the flags at least, but suffer from GNU jargonisms and usually a lack of any concise overview or example sections. Most man pages are a very short description of what the program does, and an alphabetic list of flags. For something as versatile and important as dd the description reads only "Copy a file, converting and formatting according to the operands" and there's not even one example of a full dd command given. Yes, you can figure it out from the man page, but it's like an 80s reference, not good documentation.

man pages for util-linux are my go-to example for bad documentation. Dense, require a lot of implicit knowledge of concepts, make references to 90s or 80s technology that are now neither relevant nor understandable to most users.

Plenty of other projects have typical documentation written by engineers for other engineers who already know this. man pipewire leaves you completely in the dark as to what the thing even does.

Credit to systemd, that documentation is actually comprehensive and useful.

First is Wine itself, with its implementation of Win32 APIs. I ran some games through Wine even twenty years ago but it was certainly not always possible, and usually not even easy.

Second is DXVK, which fills the main gap of Wine, namely Direct3D compatibility. Wine has long had its own implementation of D3D libraries, but it was not as performant, and more importantly it was never quite complete. You'd run into all sorts of problems because the Wine implementation differed from the Windows native D3D, and that was enough to break many gams. DXVK is a translation layer that translates D3D calls to Vulkan with excellent performance, and basically solves the problem of D3D on Linux.

Then there's the parts original to Proton itself. It applies targeted, high quality patches to Wine and DXVK to improve game compatibility, brings in a few other modules, and most importantly Proton glues it all together so it works seamlessly and with excellent UX. From the first release of Proton until recently, running Windows games through Steam took just a couple extra clicks to enable Proton for that game. And now even that isn't necessary, Proton is enabled by default so you run a game just by downloading it and launching, same exact process as on Windows.

But by now, as I also wrote in the other thread on this, ReactOS should be seen as something more like GNU Hurd. An exercise in kernel development and reverse engineering, a project that clearly requires a high level of technical skill, but long past the window of opportunity for actual adoption. If Hurd had been usable by say 1995, when Linux just got started on portability, it would have had a chance. If ReactOS had been usable ten years ago, it would also have had a chance at adoption, but now it's firmly in the "purely for engineering" space.

I understand your angle, or rather the attempt of fitting them in the same picture, somehow. However, the differences between them far surpass the similarities. There was no meaningful user-base for Unix/Hurd so to speak of compared to NT kernel. There's no real basis to assert the "kernel development" argument for both, as one was indeed a research project whereas the other one is just clean room engineering march towards replicating an existing kernel. What ReactOS needs to succeed is to become more stable and complete (on the whole, not just the kernel). Once it will be able to do that, covering the later Windows capabilities will be just a nice-to-have thing. Considering all the criticism that current version of Windows receives, switching to a stable and functional ReactOS, at least for individual use, becomes a no-brainer. Comparatively, there's nothing similar that Hurd kernel can do to get to where Linux is now.

Hurd was not a research project initially. It was a project to develop an actual, usable kernel for the GNU system, and it was supposed to be a free, copyleft replacement for the Unix kernel. ReactOS was similarly a project to make a usable and useful NT-compatible kernel, also as a free and copyleft replacement.

The key difference is that Hurd was not beholden to a particular architecture, it was free to do most things its own way as long as POSIX compatibility was achieved. ReactOS is more rigid in that it aims for compatibility with the NT implementation, including bugs, quirks and all, instead of a standard.

Both are long irrelevant to their original goals. Hurd because Linux is the dominant free Unix-like kernel (with the BSD kernel a distant second), ReactOS because the kernel it targets became a retrocomputing thing before ReactOS could reach a beta stage. And in the case of ReactOS, the secondary "whole system" goal is also irrelevant now because dozens of modern Linux distributions provide a better desktop experience than Windows 2000. Hell, Haiku is a better desktop experience.

Yet, there are still too many desktop users that, despite the wishful thinking or blaming, still haven't switched to neither Linux, nor Haiku. No mater how good Haiku or Linux distributions are, their incompatibility with the existing Windows simply disqualifies them as options for those desktop users. I bet we'll see people switching to ReactOS when it will get just stable enough, yet long before it will get as polished as either Haiku or any given quality Linux distribution.

ReactOS aims to be a system that runs Windows software and looks like Windows. But, it runs software that's compatible with WinXP (because they target the 5.1 kernel) and it looks like Windows 2000 because that's the look they're trying to recreate. Plenty of modern software people want to run doesn't run on XP. Steam doesn't run on XP. A perfectly working ReactOS would already be incompatible with what current Windows users expect.

UI wise there is the same issue. Someone used to Windows 10 or 11 would find a transition to Windows 2000 more jarring than to say Linux Mint. ReactOS is no longer a "get the UI you know" proposition, it's now "get the UI of a system from twenty five years ago, if you even used it then".

That look and feel is the easy part. That can be addressed if it's really an issue. The hard part is the compatibility (that is given by many still missing parts) and stability (the still defective parts). The targeted kernel matters, of course, but that is not set in stone. In fact, there is Windows Vista+ functionality added and written about, here: https://reactos.org/blogs/investigating-wddm although doing it properly would mean rewriting the kernel, bumping it to NT version 6.0

I'm sure there will indeed be many users that will find various ReactOS aspects jarring for as long as there are still defects, lack of polish, or dysfunction on application and kernel (drivers) level. However, considering the vast pool of Windows desktop users, it's reasonable to expect ReactOS to cover the limited needs for enough users at some point, which should turn attention into testing, polish, and funding to address anything still lacking, which then should further feed the adoption and improvement loop.

"No, people will never be switching to ReactOS. For some of the same reasons they don't switch to Linux, but stronger."

To me, this makes sense maybe for corporate world. The reasons that made them stick with Windows has less to do with familiarity or with application compatibility (given the fact that a lot of corporate infrastructure is in web applications). Yes, there must be something else that governs corporate decisions, something to do with the way corporations function, and that will most likely prevent a switch to ReactOS just as it did to Linux based distributions. But, this is exactly why I intentionally specified "for individual use" when I said "switching to a stable and functional ReactOS, at least for individual use, becomes a no-brainer". For individual use, the reason that prevented people to switch to Linux is well known, and ReactOS's reason to be was aimed exactly at that.

Sure, but that userbase also already has a way of using the NT kernel: Windows. The point is that both Hurd and ReactOS are trying to solve an interesting technical problem but lack any real reason to use rather than their alternatives that solve enough of the practical problems for most users.

With the contempt of a lot of the Win10/11 features, there's some chance it could see adoption, if that's an actual goal. But the effort is huge, and would need to be sufficient for wide desktop installs much sooner than later.

I think a couple of the Linux + WINE UI options where the underlying OS is linux, and Wine is the UI/Desktop layer on top (not too disimilar from DOS/Win9x) might also gain some traction... not to mention distros that smooth the use of WINE out for new users.

Worth mentioning a lot of WINE is reused in ReactOS, so that effort is still useful and not fully duplicated.

That's not going to happen in any way that matters. If ReactOS ever reaches Win7 compatibility, that would be at a time when Win7 is long forgotten.

The project has had a target of Windows 2000 compatibility, later changed to XP (which is a relatively minor upgrade kernel wise). Now as of 2026, ReactOS has limited USB 2.0 support and wholly lacks critical XP-level support like Wifi, NTFS or multicore CPUs. Development on the project has never been fast but somewhere around 2018 it dropped even more, just looking at the commit history there's now half the activity of a decade ago. So at current rates, it's another 5+ years away from beta level support of NT 5.0.

ReactOS actually reaching decent Win2K/XP compatibility is a long shot but still possible. Upgrading to Win7 compatibility before Win7 itself is three plus decades old, no.

The history of most FOSS projects after being running for a while.

ReactOS is mostly about the reimplementation of an older NT kernel, with a focus on driver compatibility. Their ultimate goal is to be a drop-in replacement for Windows XP such that any driver written for XP would work. That's much more relevant to industrial applications where some device is controlled by an ancient computer because the vendor originally provided drivers for NT 5.0 or 5.1 which don't work on anything modern.

> ReactOS is mostly about the reimplementation of an older NT kernel, with a focus on driver compatibility. Their ultimate goal is to be a drop-in replacement for Windows XP such that any driver written for XP would work. That's much more relevant to industrial applications where some device is controlled by an ancient computer because the vendor originally provided drivers for NT 5.0 or 5.1 which don't work on anything modern.

Fifteen years ago, they could have focused on both the industrial and consumer use cases. There were a lot of people who really didn't want to leave Windows XP in 2010-11, even just for their personal use.

Admittedly, FLOSS wasn't nearly as big of a thing back then like it is now. A larger share of GNU/Linux and BSD installs were on servers at the time, so it was a community mainly focused on commercial and industrial applications. Maybe that's what drove their focus.

I think by now ReactOS is best viewed as an enthusiast research / challenge project with no practical use, like GNU Hurd. Just as Hurd is interesting in terms of how kernels can be done, but isn't a viable candidate for practical use, ReactOS is now in the same category. Very interesting as an exercise in reimplementing NT from scratch using clean room techniques but no longer a system that has a shot at gaining any adoption.

In most of those applications, you just leave the computer be and don't touch it. In some cases (especially medical devices) you may not even be allowed to touch it for legal/compliance reasons. If the hardware dies, you most likely find the exact same machine (or something equivalent) and run the same OS - there are many scenarios where replacing the computer with something modern is not viable (lack of the correct I/O interfaces, computer is too fast, etc.)

If there were software bugs which could impact operations, they probably would have arisen during the first few years when there was a support contract. As for security issues - you lock down access and disconnect from any network with public internet access.

All that assumes that ReactOS is a perfect drop-in replacement for whatever version of Windows you are replacing, and that is probably not a good assumption.

A factory has a CNC machine delivered fifteen years ago that's been run by the same computer all along. The computer eventually gives up the ghost, the original IT guy who got the vendor's drivers and installed them on that computer with an FCKGW copy of WinXP is long gone. Asking the current IT guy, the easiest solution (in a hypothetical timeline where a usable ReactOS exists) is to take the cheapest computer available, install ReactOS, throw in drivers from the original vendor CD at the bottom of some shelf and call it a day.

LLMs know (as in have training data) everything about Kryptos. The first three messages, how they were solved including failed attempts, years of Usenet / forum messages and papers about K4, the official clues, it knows about the World Clock in Berlin, including things published in German, it can certainly write Python scripts that would replicate any viable pen-and-paper technique in milliseconds, and so on.

Yet as far as I know (though I don't actively follow K4 work), LLMs haven't produced any ideas or code useful to solving K4, let alone a solution.

My biggest complaint is that the users aren’t skeptical. They don’t even ask the LLM to verify if the answer it just generated matches the known hints from the puzzle artist. Beyond that, they don’t ask it to verify whether the decryption method actually yields the plaintext it confidently spit out.

I’m super impressed with Claude Code, though. For my use case, planning and building iOS app prototypes, it is amazing.