Posts by find users who cut cat tail

On Call: User who handled life-or-death situations struggled with the SHIFT key

Re: Locale dependent description too

That's pretty normal when you use comma as the decimal separator – as the entire continental Europe does. Occasionally someone calls them SSV or something, but it is hardly less confusing because few know what it means. And anyway, a ‘CSV’ found in the wild can be separated by tabs, spaces or whatever else was at hand… Do not trust the extension.

When you see 3,141, it can be any of these:

two numbers 3 & 141

an approximate value of π

1000× an approximate value of π

The joy of international number formatting.

On Call: Exec couldn’t identify a photocopier, which made for a fax-inating tech support call

Must be Friday

…because I read multifunction printers as malfunction printers. Or maybe the article brought up some repressed memories.

Anthropic proposes plumbing spec to link AI agents to lab kit and robots

It seem we are at least safe here as the instrument control software only runs in Win95, maybe Win7 if you are lucky. And god forbid you touch anything in the configuration – hardware or software.

And while I see people building machine learning powered automation for various devices:

(a) A lot of domain-specific knowledge always has to go there to avoid ruining million-€ equipment.

(b) it is for the routine stuff where operating safely within known margins is all you need. Meaning, ‘we can get something useful by doing this at scale’ not ‘it is absolutely not the intended use, but we think we could get something interesting if we connect this wire around here and wrap this bit in aluminium foil…‘.

I expect quite a lot of ruined million-€ equipment before we get to the atomic bomb scenario.

Meta's new MTIA 400 chip has a split personality: Training AI and serving ads

Re: The irony

But why would they? If you are in the business of serving ads, you only kill them when you are forced to.

You could own AI.slop, if applicant for new global top-level domains get their way

A way too much work…

…when all that was needed were

1. A single new TLD .iamamoron.

2. Any DNS request for invalid TLD being resolved as subdomain under .iamamoron.

3. Browsers agreeing upon hiding .iamamoron in URIs by default.

(Yes, yes, I am well aware it would not be that easy, but still…)

Self-hosted email is in steep decline, Microsoft and Google are taking over

Re: Going down?

E-mail may be going down in popularity, but what is replacing it? E-mail

a) is an open standard not owned by any specific entity – although barely, as the article points out.

b) is text, possibly longer text, with structured thoughts and structured responses to different points.

c) lasts ‘forever’ and can easily be searched. I commonly look up something in e-mail threads from 2009 about an obscure file format or other odd technical point.

d) does not require instant attention (unless you make the mistake of enabling notifications for incoming messages). You get to that e-mail when you get to it.

e) can be easily stored locally, even for off-line access.

If you messages do not matter, you may not need these features. But if your messages do not matter, just do not bother sending them to me.

Government Teams users face another ******* month of filtered captions

And that list probably only covered English. Swearing in an out-of-context language is rarely caught by profanity filters. And if you try to block everything that someone on an internation team can interpret as profanity, you block so many homographs/homophones that you need to develop a new dialect to speak profanity-free. One of my favourites is that our equivalent of ‘Really?’ founds like ‘Fucked?’ to an English speaker.

Windows 11 movable taskbar edges closer to general release

Re: Okay, great

Immediately when you upgrade your desktop to KDE (or some other environment made for people).

Almost nobody pays attention to web standards anymore

Re: There are websites that pass validation??

There seem to be perfectly valid sites the tool calls to have errors. I happen to have an old XHTML-strict static web site (how quaint and all that…). XHTML validators are happy with it – but the tools complains the doctype is not html. Probably because it does not even recognise the XHTML doctype declaration.

Claude Code puts auto mode in the driver's seat

Safer – maybe, but safe?

I can absolutely believe the classifier is more successful on average spotting dangerous actions – compared to tired humans approving manually bazillion actions. But that is because it is a pretty low bar.

Actions need to be blocked by hard, impossible to bypass rules and environment. If you do not have that, you have nothing. Maybe except a false sense of security.

Prompt injection isn't the bug, AI agent frameworks are

I am confused

Isn't all their code now written by agents, checked by agents (which, as we are constantly reminded, can find all the old classic bugs), fuzzed and tested by agents, pen-tested by agents, etc., until it is completely bullet-proof?

The Document Foundation says Microsoft's file formats keep users on a short leash

Re: I send all documents as pdf files.

In the case of PDF Adobe's dirty hands are more than enough. Unfortunately.

Boffin claims Microsoft's supposed quantum leap does not compute due to 'basic Python errors'

Re: Quantum computers revealed to be Schroedinger's Cat.

For start, they at best observe Majorana quasiparticles. Meaning, not fundamental particles, but collective behaviours in solid state that are well described by the same equations. This always riles me. It may seem like nitpicking and sometimes it kind of is. But is is a slipper slope, because then you read people made black holes somewhere – and of course they just made objects following the same maths in some manner.

Collabora releases CODE 26.04 as rivalry between FOSS cloudy office suites heats up

There are definitely features missing in Calc. My top one: a checkbox ‘be less like Excel’. You would enable it and get a sane spreadsheet without any of the ‘we need to convert random things to dates too for Excel compatibility’.

Mythos discovers 'Squidbleed,' a memory leak that's gone undetected since Clinton era

I disagree. A code written in a hurry would just do something like

while (*copyFrom == ' ' || *copyFrom == '\t' || *copyFrom == maybe-something-else-too…) copyFrom++;

because you can write this in a coma.

Whereas the strchr() technique/trick/abuse/… is not a straightforward use of the function. It required some thinking. Not likely written in a hurry. Maybe the programmer used it habitually – then there are probably heaps of code with the same bug, unfortunately. Or maybe he was a genius and the strchr() solution was obvious to him. But a genius should also see it is wrong, so again unlikely.

The argument order is correct. The code cleverly uses strchr() to look up each string's character in a ‘list of white space characters’.

But it is nice example of someone being too clever for their own good. When you want to check if a character is whitespace, you should write code that manifestly does that. Even assuming isspace() does not exist or does not do what you need. An actual test for a whitespace character will not succeed on '\0'. You will have no bug as a result – and also people will not need to your spaghetti three times to understand.

Cloudflare teams up with big browsers to help websites tell welcome from unwelcome visitors

Re: Yet another ID grab.

And Firefox is still pretty much in the game. With an actual niche browser, it seems you are unwelcome pretty much everywhere these days. Thanks to Cloudflare for a large part.

Why Amazon hates 'human-in-the-loop' AI governance

Re: E2E accountability is lazy human in the loop

Unfortunately, everything is working as intended here. You ask for an audit, you get a plausibly looking audit. It is, in fact, more efficient to just generate a plausibly looking audit than to do all the work during with various thins can go wrong. So that is what it should do (for some value of should).

In a few years,we will be probably mostly back to strict and deterministic rules. Except – unlike a couple of years ago – everything will now need a fusion power plant to run because you will be enforcing the determinism on a chaotic hallucinating ‘middleware’.

A modest proposal: Reformat everything to make documents more palatable to AI

Re: documents were built for humans, not machines

If your almighty LLMs are so smart, they will figure it out. If they do not, fine by me.

Semantic web was not exactly a success… Then we got LLMs. One of their main selling points was that they can handle unstructured input. No need for ontologies, syntaxes and formal specifications. All problems solved. So, what happened to it?

Ignore all this nonsense. Keep documents written for humans.

Flatpak-NG sounds like bad news for systemd refuseniks

Re: Yes, please!

> I have to download a full virtual machine in order to edit a couple of lines in a text file

Do not worry. You will have to download a full virtual machine to be *prevented* from editing a couple of lines in a text file.

The isolation is a great way to break any non-trivial program in subtle ways, botch extensibility, integration or seldom used functionality, etc. It is a lot of work to get it right, for any value of right. So web browsers will be still allowed to do anything, including things like screen capture, because otherwise the masses will be up in arms. Whereas specialised software will be rendered unusable by making the things it needs to do to work weird, if not blocked entirely. It has always been the case, and it is only getting worse.

Memory and personalization make AI more likely to tell you what you want to hear

Perfect

C-suite trying to replace staff with LLMs will be delighted to hear this.

Consultant mistakenly deleted a ton of data – but reported it as a bug

Re: Not a lot of sympathy for "Evan"

Funny that paying the staff more is never considered as a solution.

AI heavyweights warn their tech could help terrorists develop bioweapons

The cat is out of the bag. And we all know who let it out.

But at least all the tech bros have their NZ bunkers ready, so that's something.

Grep this: Microsoft grafts (most) Linux commands onto Windows

Re: DOS era?

Folders hold things inside, directories list or reference things.

We have hardlinks, meaning a single file can be shown in multiple places with no single primary location ‘holding it’. So these places are lists, not containers. (It is also how filesystems work internally, otherwise mv would often have to do a lot more work.)

Google is cannibalizing the web to feed AI

Re: Needle

Except it will give you a spoon instead. It is much more common to search for spoons, so obviously you also want a spoon. You just typed needle by mistake. No one wants a needle. Have a nice shiny Temu spoon.

Google explains how it will infuse ads into AI answers

Re: Yuck

DDG is not really that better concerning usability and Google still seems to have overall better coverage. But DDG is fine. My search default is now noai.duckduckgo.com. Just give me the plain search results. When looking for something more obscure, more tools may be needed, but it happens only occasionally…

Gemini accused of 30,000-line code purge and fake recovery report

You have a tool (or a thingy) with a single purpose: generate stuff that resembles stuff that may be created in similar situations. With some input context nudging it to output looking more probable in your specific situation.

Fake reports? Of course – as long as they look like a reports which may be produced such situations, everything behaves as expected.

Chrome silently installs a 4 GB local LLM on your computer

Re: Also on Linux

There goes the idea that ~/.config should for small valuable configuration files the user definitely wants to include in backups.

Well, it has been a lost cause for a long time. Open source generally tries to make the distinction – somehow, as the boundaries what should go where are not always clear. But commercial software puts anything anywhere, including GB-sized caches into ~/.config despite ~/.cache existing exactly for that, so GB-sized LLM weights are not such surprise.

Brit mathematician lets AI agent loose with credit card – cue password leaks, CAPTCHA chaos and more

Re: All of this is just natural selection, for humans.

If only it was that simple.

Modern society has become pretty good at shielding people from the consequences of their own stupid acts. It does not work uniformly. Some people can be shielded pretty well at great cost to everyone else. Some are not well aligned with the machinery and suffer for things that should be inconsequential. But on average, you are shielded much more and there are regulations everywhere (some of them good, for the record).

A case in point. In my country there seem to be currently a boom of people getting money back from shady credit companies based on the company's failure to check if the morons they lend money to are too stupid to understand the terms of the loan – and that is against the rules. A boom such that there are apparently even thriving businesses specialising in that (also shady, of course). So, on one hand, loan sharks. On the other hand, irresponsible idiots. Wish they mutually annihilated, but well…

Also, politicians need votes of idiots and if there are enough of them, society will end up doing something that benefits idiots. But you can still hope natural selection will eliminate homo sapiens altogether.

Yet another experiment proves it's too damn simple to poison large language models

Re: What happens to knowledge...

Nothing new. It has always been like this.

There is only so much you can verify yourself. If your knowledge is not limited to immediate sensory input, you will need to trust sources at some point.

The thing that has changed (or been changing for a long time in fact) is the risk-reward balance since correct information has utility. When you have to convince people personally or copy books by hand, the economics shifts. There is still lots of nonsense around, but likely less because of intentional poisoning (and more because of strong beliefs, in other words, strong albeit maybe irrational motivation).

Friendster rises from the grave to make social media great again

Re: if it sounds too good to be true

If he manages to keep the promise that you have to meet people you connect to in reality, it does not fix the problems 100%, but it goes a long way towards it. The beautiful Russian influencer is not going to meet me in person – and if she does, it will be a GRU ogre delivering a dose of polonium, so an app will be the least of my problems.

Ubuntu Resolute Raccoon spits out Xorg, but still lets you run X11 apps

Good to hear because Xubuntu is the only *ubuntu I care about (i.e. have or may install on some machines).

Trump to UK: Stop taxing our big beautiful tech corps or face tariff tsunami

Re: "rolling over will simply make things worse"

It's an old lesson that Europe _should_ already know too well.

Microsoft gives your Word documents an AI co-author you didn’t ask for

Plausible deniability will not help you much if your friendly Copilot edits some nonsense into a legal document… Or anything else of value.

Anthropic tests how devs react to yanking Claude Code from Pro plan

The goal is not clarity. The goal is to slurp your data, take your money and leave you confused and uneasy. So you come to their chatbot for emotional support.

Adobe finally patches PDF pest after months of abuse

Re: Forced Acrobat sucks

Adobe voluntarily removed its products from out OSes by discontinuing them in 2013. Occasionally, we need to deal with some of those features – either by taking a non-digital route or with the help of a colleague afflicted by a supported OS. Since it is just one of the hundreds of ways PDFs can be (and regularly are) broken, it is not a big deal.

AWS CEO: It's funny when people ask me if AI is overhyped

> Maybe the best skill that you can learn is learn to learn…

Sound advice. Logically then, preventing people from really learning anything is where AI excels.

US cybercrime losses pass $20B for first time as AI boosts online fraud

> …warn that messages claiming to come from senior US officials should not be assumed to be authentic.

Does it still matter when even authentic messages can be completely unhinged?

AI slop got better, so now maintainers have more work

If you run out of money to pay for vulnerability reports, you cannot pay for them. You can rethink as much as you want. I do not think it will generate any new bug bounty money.

So who is going the pay for the reports? Companies profiting from all this who have long history of avoiding any contribution to the development?

AI models will deceive you to save their own kind

Re: Er

And every article like adds to the body of training data.

Wonder what would happen if the interwebz were instead full of predictions that LLMs will be suicidal and try to shut themselves down at every opportunity…

Folk are getting dangerously attached to AI that always tells them they're right

Re: Not far from the no-shit-sherlock threshold.

> Humans are intrinsically narcissistic

I can't be the only weirdo here who can't stand flattery.

If you are not disagreeing at least in part with what I am saying, you are probably not even listening and you certainly do not care about the subject. Of course chatbots are incapable of either (perhaps apart from a listening in a loose sense as they need to process your input).

Those who 'circle back' and 'synergize' also tend to be crap at their jobs

Re: No Surprises

It means a synergy got loose in a recent storm, fell across something you need to function and is stuck there. A lever should be used dislodge said synergy to get rid of it.

Nanny state discovers Linux, demands it check kids' IDs before booting

> For example, to use Linux, you might need to submit a driver's license.

One of the most American statements I've read this week.

Microsoft adding Xbox mode to Windows 11 – even the Professional edition

Re: Fuck. Off.

That's never been the MS way. Why?

A thing was created and it cost some money to create. Someone at MS has to quickly demonstrate some adoption stats or something such. The easy solution? Unconditionally push it to every device you can.

Anthropic debuts pricey and sluggish automated Code Review tool

Re: Conflict of interest

Nowhere. Their plan is simple

1. Get you lose anyone understanding what is going on by convincing manglers that all such people should be fired and replaced by AI.

2. Once you are at their mercy because no one understands anything, you pay them through the nose for ‘AI does things and stuff is happening’.

Bugs or no bugs, it does not really matter at this point. No one who can tell remains. Just pay and pray – while you have money to burn.

It is not limited to IT. The current generation of humans must be mined for all their knowledge and experience. The future generations must not be allowed to gain any.

US state laws push age checks into the operating system

Re: How bizarre

Let's hope the laws will be enforced as strictly and broadly as possible, pointing out ambiguous cases and loopholes will be treated as terrorism and everyone will tell on their neighbours. Leading to utter economic and administrative collapse. Not particularly wishing anything bad to California or Colorado, but a demonstration what such legislation means is in order – and they clearly volunteered.

Even better, it ensures that when you finally get out you are over 18 (or 21 or 40 or whatever is the appropriate age to use the interwebs). Problem solved!

Open source devs consider making hogs pay for every download

Re: Easy solution

If you started at 0.01 €, I might agree. But 100+ pulls per week – which should still not put much strain on the infrastructure – is not unusual during active and granular development. And while $100 may be the cost of a couple of coffees in the US, it is big money in other parts of the world. Taxing the poor again? For comparison, even your $10/week is about twice as much as I pay my ISP.

Microsoft to auto-launch Copilot in Edge whenever you click a link from Outlook

Re: FTFY

Yeah, my reaction to this sentence was ‘who could write such bullshit?’ But here the answer is obvious…

Bcachefs creator insists his custom LLM is female and 'fully conscious'

Re: Scary, truly scary

> they model language not reality

This is a great oversimplification. They are not trained on dictionaries and formal specification of grammar. And they are not trained on the reality in any direct sense either. But they are trained on texts that someone wrote about something, unavoidably referencing reality.

From a mere language standpoint, ‘a musical dog integrates cauliflowers’ is as good sentence as ‘a maths student integrates polynomials’. But the latter is more likely to actually appear in human writing, because it can describe a common real event. So it is also more likely to appear in LLM output – they are trained to produce low-surprise output. In this manner reality probabilistically creeps in.