I can't say there's a single thing I enjoy about any of this any more. I think I understand why so many IT people retire early and take up farming.

I suspect it's going to be important but I wonder if these companies that are demanding it are being premature.

I remember when syntax highlighting came in. For a good while I disabled it as I found it distracting but over time I got used to it and now I prefer to have it on (although I'm still not greatly inconvenienced if I don't have colour)

I remember when LSPs became available. For a while I didn't bother but now I find it hard to work without one configured (in vim) although I do still resort to grep/sed etc, especially where the LSP doesn't work in some C++ where type deduction is hard.

And now LLMs. I use them via a web prompt and they can save considerable time where I'm looking something up. But they've also cost me considerable time where I've relied on the answer which, once I actually spent the five minutes reading the (long) manpage, was so obviously wrong that I'd be concerned about any programmers long term success and I'd certainly be trying to limit their access to any code that I'm responsible for!

I think more than anything it's this utterly absurd wrongness that they do with such verisimilitude that make them dangerous currently.

does unlink in debugfs update the free block list?

Yes, when you use unlink on a file in debugfs (or via any means in the Linux filesystem stack), it does eventually update the free block list -- but not immediately at the debugfs tool level unless explicitly committed or updated.

In debugfs specifically:

        debugfs is a low-level filesystem debugger. It does not mount the filesystem or execute actions in real time the way the kernel does during normal operations.

        If you run the unlink command in debugfs, it removes the directory entry, and depending on the inode's link count, it may also release the inode.

        But to update the free block list, debugfs must explicitly free the inode and associated blocks.

In recent versions of debugfs:

        unlink can update the block and inode bitmaps if used correctly and if the file has no other hard links.

        But sometimes manual intervention (e.g., using clri to clear inodes or freeb to free blocks) is necessary in corrupted filesystems or after a crash.

Summary:

        Yes, unlink in debugfs can update the free block list.

        But this happens only if the file is fully unlinked (no hard links remain).

        Remember that debugfs works outside the kernel, so changes are made directly to disk structures -- use with caution.

But:

man debugfs

              unlink pathname
                            Remove the link specified by pathname to an inode. Note this does not adjust the inode reference counts.

I really don't get it. Isn't AI supposed to save everybody time? Someone has gone to the trouble of describing their issue to AI in sufficient detail that the AI can create an issue for a human to understand it and do the work, so why doesn't AI just create the pull request?

Even better, if the project maintainer doesn't like it, the requester can for k the project and have their feature/bugfix/whatever and the world is a better place.

When I create a bug (even a wishlist bug) I often/usually have a patch too, and I'm usually running with that patch already because the reason I spent the effort on the bug is because I needed it fixed.

Sadly, this is probably true. I still use my phone as a communications device but it's slowly getting harder and harder to use like that. Google (and Apple although I've not used it much other than to try when Google first broke always on VPN and discovered that Apple has the same problem without the option of a custom rom) are focused on tracking and advertisment delivery (AKA money extraction).

I use always on VPN (to a private server), that way there's no information in my IP about where I am, everything comes from a handful of IPs. Sure, websites etc can (probably) know it's me each time but the don't get any information about where I might be.

But both apple and google have (IMO deliberately) broken this. I don't know of any way to have always on VPN working on Apple and receive things like whatsapp because the phone going to sleep disables the vpn. On Android it's still possible[1] on a rooted device with a custom rom but you do have to accept that a significant proportion of apps won't work on a rooted phone - so you end up needing a second phone to use the first phone as a hotspot.

[1] I suspect that one day Whatsapp will refuse to run on a rooted phone.

I guess their model goes:
1. Buy established business
2. Replace lots of staff with AI
3. Sell at profit.
4. Who cares what happens to the business long term

So this lets non-US investors access US stocks.

And when the orange menace decides he doesn't like it and forcibly converts them to the underlying stock but doesn't allow the investor to sell their holding in the US they're stuffed, much like happened to ADR holders of Russian equities when Putin invaded Ukraine and Putin decided to require ADRs to be delisted and converted to the underlying share.

Oh, this is crypto, there is no "underlying share", you just think there is.

I work for a (regulated) american company but am based in the EU.

One of the problem is the different way regulations are written and enforced.

In America the regulator comes in, says "You need to do X, Y, Z to get into compliance". We do X, Y, Z and next year the regular is "all good".

In the EU the regulator comes in and says "These things aren't right. X, Y, Z must to be done". We do X, Y, Z and then next year the regulator is "This isn't good enough, That previous report wasn't a list of things to do, it was a list of things that definitely weren't right".

This makes it much harder, and much more expensive to comply. Not least because there's a constant conflict between "is this enough?" and "should we spend more money to do more?".

We now have a good relationship with the regulator but it was a painful few years.

I don't know much about the asia side but I think the regulations are more strict, easy to accidentally violate when a bug gets introduced, but easier to understand and know when you are and aren't in compliance - therefore at least in our company they're considered easier and cheaper to deal with.

Hasn't this just reset to before the spat started in March?

I suspect the biggest problem the US is going to have now is restarting the supply pipelines from China. As the adafruit example showed, you can be hit by tariffs that didn't exist when you placed your order.

So a) there's going to be some importers unwilling to place orders even though the tariffs are back to what they were and b) the ones who do place orders are only going to be delivered in 30+ days. The current China->US pipeline is very significantly down in volume (and sailings) although the early stockpiling has hidden that from the US consumer so far.

https://nypost.com/2025/05/09/...

https://www.cnbc.com/2025/04/2...

I'm sort of amazed that companies like Microsoft, Google have 30% of their codebase written by AI.

It's pretty much guaranteed that they'll have been training on GPL2 and GPL3 code and it's going to be really hard for someone to tell if it's copied from copyright sources.

Yesterday I needed to add lseek support to a fuse driver. I asked chatgpt for a basic example and it was completely obvious that it had lifted it (maybe modified) from an open source driver.
(It came from here https://lists.nongnu.org/archi..., the code was sufficiently similar that with just two keywords copied from the stuff that chatgpt wrote, I got that code as the top hit in ddg. And the chatgpt code looks almost identical other than having removed the comments - I haven't checked every line character by character)

In this particular example of lseek support in a fuse driver, it's not the sort of thing that you can cut and paste, so what the precise licence of the original code was really doesn't matter, all I wanted was the basic framework and what codes to return on error and the guts I implemented from scratch. But I'd guess that it will take far longer for programmers to research where AI generated code came from and whether it's going to "contaminate" a code base than to write the code.

AI is a great help. it took 30 seconds for me to get an outline implementation, finding an opensource implementation, finding the code, discovering that it's mindbendingly complex and looking for another simpler example, could easily have taken me 30 minutes or more, particularly as fuse drivers is not something I've really worked on before.

I'd think they'll also be worried about the EU deciding that they cannot afford to use an American based cloud company for government business. (I'm assuming that Microsoft is used extensively)

It's sort of like what is happening with defence. The EU is ramping up its plans and spending but it's simultaneously planning how to limit its dependence on the US.

https://www.theguardian.com/po...

Of course it could. The more interesting question is how it will be acknowledged, by mathematicians, by the AI interests, and by the media.

As with many other fields, computers offer the opportunity to search vast mathematical spaces that are beyond the ability of humans to investigate.

But the as yet unanswered question is how to tell a computer to look for something "interesting". In searches for drugs, for example, we might ask the computer to find a drug that will bind to a particular site or otherwise disrupt a metabolic pathway, but ask it to find *new* interesting ideas is beyond description, let alone execution.

This isn't particularly surprising, it's a vanishingly rare skill even amongst humans skilled in the field.

And so it will be with AI proofs. There will be great fanfare the first time an AI proves, with or without guidance from top mathematicians, some previously unsolved conjecture. "ChatGPT proves the Goldbach Conjecture! AI surpasses human intelligence!" but assuming that the Goldbach conjecture can be proved using already known mathematics then all the work is done, and all that remains is searching, something that AI is particularly good at. The search space is vast however, and it's just as likely that the only way an AI can find a proof is with the intuition of genius mathematicians stopping it going off down unproductive routes. The AI will find the proof but it may, or may not be something it could have done alone.

And, should it happen, it's quite likely that the AI proof will spur countless interesting discoveries by the mathematicians skilled in the field once they see the "trick" to solving that problem. Possibly half a dozen other unsolved problems might fall, not due to AI but due to *understanding*.

The second thing that will happen, but probably will slip most of us by because it won't make the mainstream press and we don't read mathematics journals, is some deep and profound new conjectures (with or without proof) that will be mostly the work of mathematicians but with a good dose of AI assist. This may, or may not, get credited to AI but will involve the sort of intelligent searching that AI can deliver on a scale that isn't possible by humans due to the limited number of them, their low speed relative to computers, and their propensity to make mistakes, particularly when doing boring grunt work.

The final thing that probably will happen eventually but doesn't appear to be on the horizon, is AIs coming up with new and profound conjectures absent the guidance of mathematicians. This is going to be hard to identify, not least because there's a vast difference between a computer coming up with 200 "conjectures", 198 of which are "completely wrong headed", 1 of which is already known and 1 of which the mathematicians looking at its output say "Ohhh, that's interesting" and an AI coming up with 3 conjectures, 2 of which are already known/proved and one is new, novel and interesting.

Jesus,

They checked healthy arteries.
They checked diseased (clogged) arteries and found 16x the level of microplastics of the healthy arteries
They checked arteries of stroke victims and found 51x the level of the healthy arteries, 3x the level in the clogged arteries.

And this is homeopathy?

Microplastics aren't at all hard to detect with the right equipment, they're just hard to see without. It's like cholera in early Victorian times, we're at the level of knowing there's something but not understanding exactly how or why it a problem or why it impacts some people far worse than others.

You're the doctors saying "it's a miasma", these are the doctors saying "the miasma theory makes no sense, how come the nurses who care for cholera sufferers are less likely to catch it than people living at the other end of the street".

Nobody knows why plutonium is dangerous to ingest in "homeopathic doses" either. It's a heavy metal, it's radioactive, and it bioaccumulates - it's a really good idea to avoid even low levels of ingestion, and a really good idea to have international regulations controlling how plutonium is produced, stored, and disposed of (there are isotopes of plutonium that aren't useful for fission) even though we don't actually know how it will kill you - the primary theories are its heavy metal toxicity and the effects of its radiation on the bone marrow.

And the same with microplastics, they're not heavy metals, they're not radioactive, but they do cross the blood-brain barrier (I don't think plutonium does) and it appears they do bioaccumulate, possibly in the worst places in the body and they might have a genetic level impact on cells. AFAICT this study is looking for correlation rather than causation although they suggest a possible causative effect, I'm not sure if that's speculation or an informed guess - you would need to be an expert in the field to know and I'm not.

One can't eat eggs with a silver spoon so one needs gold spoons for eggs, aluminium being so common and cheap today. We need to reserve the hallowed halls of learning for the sons of Lords who know their superiority.

The hoi polloi today, don't even know why they must respect their betters.

Truss has worked diligently to earn comparisons to Trump

Truss came within a hairs breadth of bankrupting the UK. It doesn't get a huge amount of noise outside of the specialist press but many UK pension funds were highly leveraged using LDI. The huge spike in the bond rate caused many funds to be dumping equities at any price in order to meet margin calls due at the end of the day.

You can argue that the use of LDI, and the leveraging was bad and the pension funds should not have been doing it, but if Truss thought that then she was in a position to reign this in - she had a substantial majority in parliament and could easily have changed the rules for pension funds. But (and I don't believe for one minute that she had thought any of this through), to bankrupt the pension schemes because you don't like the way they're using LDI so that the PPF (Pension Protection Fund) will have to pick this up is insanity of the highest order. I'm not actually sure what happens if the PPF goes bust, it's not something that seems to have been considered - dealing with an individual fund failure can be covered by future levies, but the sort of failure that was imminent would have meant that the remaining firms would also have been forced into liquidation if the levy had been asked to cover the new liabilities.

I understand that the US doesn't have the same level of defined benefit pensions (currently, around 1 in 8 UK people either receive or will be eligible for a DB pension) and I would assume much less regulation of the pensions and little if any statutory protection if one goes bust, so the precise set of events that brought down Truss could not happen in the US. But a dedollarisation that sends US Treasury yields soaring the way the UK yields went is still bound to have catastrophic side effects. Trump's efforts are weakening the dollar but not crashing it and there are some good arguments why a weaker dollar could be good for the US so, as of yet, Trump is not emulating Truss although it's not inconceivable that he could in the future.

I often no in my mind, but my fingers do the typing and dont always get it write.

Just say know.

Woosh.

Consider: the missing apostrophe in "dont" and the wrong form of "write" which should be "right".

Impressive! I literally cannot tell whether the anonymous coward that wrote the woosh is an imbecile or a genius.

Damn missed one:

It wasn't given enough time.