Posts by that one in the corner

Humanoid robots are still novelty acts, but investment is surging to make them real tomorrow

Re: God would have given all of us two digits instead of five with opposable thumbs

Needs to be robust for when they put out all those forks in the road.

God would have given all of us two digits instead of five with opposable thumbs

This may be part of his problem: he hasn't figured out that in industry you can design both the gripper *and* the thing being gripped in parallel.

A pair of parallel pincers has been a working solution for, ooh, a century or so. Heck, a lot of the time, you don't even need to have the pincers move relative to each other: use a standardised palette and the fork lift[1] can handle everything you need it to. If you don't want the stuff to be in a palette, you just pop handles in each side. Advancements let you point the prongs in various directions, have them in all sizes, let them run longer macros[2] and react to unexpected things in the environment.

[1] a fork lift is a perfectly good industrial robot, more so nowadays with teleoperated units.

[2] macros used to be dine by moving the cams

North American air defense troops ready for 70th year of Santa tracking

Re: NORAD need more computing power

Why are you letting the sheep eat humbugs? They should stick to baahley sugar canes.

Pen testers accused of 'blackmail' after reporting Eurostar chatbot flaws

Pass on details of flaw, be accused of bad faith actions such as blackmail

So we've turned the clock back to the early days of companies being online, when every report of a flaw in the public facing system was met with accusations of "hacking", including claims the "hacker" (aka pen tester nowadays) was trying to extort or blackmail the company.

And probably for the same reason: the company[1] thinks its system *must* be perfect, so the only way it can go wrong is by deliberate attacks.

[1] The senior management, that is; everyone else knows it is proped up matchsticks.

NASA tries Curiosity rover's Mastcam to work out where MAVEN might be

Extraplanetary

Peek-a-boo

Microsoft wants to replace its entire C and C++ codebase, perhaps by 2030

> Then lesser coders came, and we got printf and the rest

Um, printf() is one of The Original C Standard Library Functions (see K&R 1st edition). The lack of type data that any variadic C routine suffers from (and C+ necessarily inherited) is one of Life's Great Blunders, from our lofty perspective, but back then...

Stopping your incoming lesser coders from adding new printf()s into your existing code is always a problem, but you can get ahead of them by

#define printf PrintF

Int PrintF(const string& str, const P& p1 = P::Nowt(), const P& p2 = P::Nowt(),

and fill in as many p1, p2 ... p100 as you feel the need for. Define class P appropriately to deal with all the standard types etc etc, you know the drill.

Re: Microsoft should improve its recruting procedures...

> rewrite object-oriented C++ code into a non-object oriented language

You get a copy of cfront (i.e. the original implementation of C++, with all the object-orientation but a bit lacking in the template department) and use that to compile the C++ into plain old C.

cfront may(!) need a tweak to the grammar it accepts, but overall that is a doable thing and the output wasn't spaghetti. Lots of weird names for temporaries and you will see the unvarnished mangled names, but once you get the hang of the mangling algorithm...

US punishes China’s ‘dominance’ of legacy chips with zero percent tariffs

Re: The result of an investigation into China’s semiconductor industry

The point that is the "cost of living" to "wages" ratio[1]

So saying "I doubt you could keep slaves alive in the UK on Chinese factory worker pay" misses the point, because you are effectively mixing units. Ditto "Chinese factory worker pay in the US would effectively be slave labour". For the day-to-day you have consider wages against *infranational* trade.

The question of selling cheap Chinese chips into the US is about profitable *international* trade, and the skew in currency pricing on the trading floors.

So, instead of something like

> A better plan (for the USA) would be to focus on the education required to make high margin products

an alternative plan would be simply to devalue the USD against the Yuan. Say by making the US Government mercurial, so general international trade falters; scaring off tourists... So much simpler than actually having affordable higher education. And given the videos of preachers calling for the removal of women's suffrage and similar reversals back to the Great Days of the Puritans...

[1] and whilst there are definitely Chinese at the low end, how does that actually compare to the US? With multiple jobs that only work if the tips are good because the pay isn't; and the number of homeless versus low pay factory workers... I do not know reliable numbers.

The result of an investigation into China’s semiconductor industry

Shouldn't that have been

"the result of an investigation into the USA’s semiconductor industry" and why they stopped bothering to make these mature devices[1] if they are so damn important to the USA.

> sweeping non-market policies and practices

I.e. being better at it than the USA, who fetishised "new and shiny, with bigger markups" and "you MUST grow and grow and grow, we will sue you for not maximising fiduciary responsibility if you even *try* to just keep making these old-fashioned parts at rates that let you break-even, pay a consistent but smaller dividend and tie up any excess in planning to keep running in ten years time".

[1] I utterly refuse to condone the idiotic usage of "legacy" in this fashion; if enough of us just laugh in the face of marketers and salesdroids when they try it, maybe we can get them to stop.

Windows is testing a new, wider Run dialog box. Here’s how to try it

The IT chap at my last job used it a lot: some tweak was needed on your PC for whatever arcane reason. A nice lad, overall, but:

He tended to scowl when at my desk, because of the lack of a "Windows" key on my beloved keyboard.

And looked blankly when I just alt-tabbed to the inevitably already open cmd.exe window; maybe ther *is* some oddity in the execution environment between the "Run" dialogue and a generic cmd.exe session, but I never found out what it is...

How to enable the new Run dialog box

> Launch the command line

And just use that.

Minimise it and keep it around, it'll be useful again soon, and with a useful history...

Memory is running out, and so are excuses for software bloat

Re: Anybody remember the linker?

> The linker can and indeed does remove unused functions and static variables

Depends upon your linker having that feature; for common targets and/or Big Project toolchains, yay, great. But you can't guarantee that, even nowadays, especially for oddball targets that are more likely to be memory constrained. The plan I outlined works for every target - *and* would have done so from the Before Days when we were still calling them "Link Editors" and worried every time over how well they'd enable the Loader to do its job.

Ok, yes, w.r.t. the core argument of TFA (wanting better memory management on "standard platforms", i.e. Windows and Linux on PCs and servers) you are more than likely to have a pleasingly capable linker (and compiler willing to work nicely to allow the linker to do its very best job).

> Possibly because they're being charged by the second for build resources, either directly or via that curse of the corporation, an "internal market".

And when you try pointing out what happens when you compile & link a release copy just the once, then run it thousands, millions of time... (Ah, "dev" and "factory floor" are two different internal cost-centres and I only have to budget for one of them").

Oh, for the simpler days of 2007.

(Of course, *I* always use that time to plan out module tests. Ahem)

I'm looking down the barrel of a running a "Learn a bit about MCUs" course

I am tempted to stick with C or C/C++ 'cos you (needn't) waste any space or execution cycles and I could use the good-old Atmel MCUs[1] or to use some larger devices and start them off with Python[2].

The current pundits tell me to use Python, it is easier for the little darlings. But, as did others here, I started off with 256 bytes[3] in My First Micro and still do things like print out sizeof(MyClass) for everyting in "my collected personal favourites" library, weeping if they are forced to grow larger. Seeing the space needed by - and the slow execution speed of - Python just jangles the nerves.

But if I can start them with Thonny and not worry about "this is what a compiler & linker & uploader do, you'll need to know this to understand the error messages" it will be easier.

But the bit-packed C structs can have a 1:1 with the hardware and are efficient. You can fit data for more RGB LEDs.

But the Python-capable MCUs have enough RAM to cope with any beginner's needs (AND the slippery slope begins).

But... Python better. But... C better. But... But...

(Exit stage left, carrying table whilst banging head on it).

[1] in other words, do a "Intro to Arduino" but try desperately to ignore all the daftness on recent "Arduino".

[2] CircuitPython or "plain" MicroPython, either one. Or swap between them, keep everyone on their toes.

[3] yes, I said bytes, not Kilobytes (or kibibytes), not Megabytes...

Re: Anybody remember the linker?

As noted above, linking to libraries as DLLs rather knackers that ideal.

If you are prepared to go for static linking (which has many benefits, btw, such as being certain which copy of routine x() you are calling[1]) then the prevalence of big, complicated, libraries also kicks the ideal in the nadgers because very few library authors are willing to break the code up into the thousands of separate dot-c files necessary for the trick to work[2]. Because it is a lot easier to deal with "only" a few hundred source files than many thousands, and a lot easier if you can load one source file into the editor and see all the methods that make up a single class, for example.

All of which is a dreadful shame: all it comes down to is how the source editor works, as they all still have you think of terms of "which file did I put that routine in" and "keep related things together for life cality of reference". But sorting out the editor could have been done decades ago[5]. The version control eould have followed easily (e.g. take the same manifest as the Linker and don't bother the User with the fact that so many little files are involved). A simple(!) tool to break down existing full-fat files[6] to get things started. There *would* also have to be some fiddling with the programming languages, but nothing too drastic: when the tool breaks down your full-fat files, it will have to add some annotation to indicate that *this* static declaration is visible *here* and *here* but not *there*; oh, and "there" has its own static (in the C/C++ meaning of the word) with the same name[7]...

Anyway, bottom line, we *could* have had (still could, really) a toolset that FORCED the ability to do a minimal static link. But who cared? Cared enough to fill in the gaps I've left, write the tools - and then who cared enough to run these tools in their build? Personally, I'd love to *have* such tools, but, um, well, haven't got to grips with creating them (see note about it could be done, now, with LLVM, but that wasn't around when...) but everybody else is happy with DLLs...

BTW there ARE arguments for going in exactly the opposite direction: for example, although you *can* happily build SQLite as "one dot-c per compilation unit" the documentation makes it plain that they prefer you use a "all the dot-c files pulled into one, single, humongous compilation unit"! Which will totally defeat the Linker (one reference to any part of SQLite and *everything* gets dragged in). BUT they then claim that doing this lets the compiler's optimiser have full reign over the code and this makes the result faster and betterer, so it is worth doing things this way. Whicn is true, you *can* get better speed, if not space, optimised code with this trick. Sometimes. Don't just think it'll work for every bit of code you have...

[1] and others will say this us a DISadvantage, as you can't update x.dll yourself with a fixed copy of x() when the vendor goes byebyes; this - discussion - can rage on and on...

[2] because the Link Phase still does what it always did: pull in the separate, but entire, dot-obj[3] files from within the dot-lib[4]

[3] sorry for the DOS'isms, replace dot-obj with dot-o etc to fit your preferred OS; and dot-c with your preferred programming language's default file extension.

[4] because a dot-lib file really is nothing more than an archive file of a pile of dot-obj files, where each dot-obj is the result of compiling a single compilation unit (which is, in the simple case, one dot-c file per compilation unit); this is made obvious when you spot that the "make this lot into a library" command may literally be the 'ar' file archiver command. Like zip files, only an older, simpler, format.

[5] E.g. an agreement on a simple text file - call it a "manifest" - that the tools kept up to date and prevented you from editing (e.g. put a checksum, in text, at the end - you can read these manually, just not edit rhem and expect to have them work) which listed all the files that the editor had used to store all the separate bits of code: putting it crudely (note that - being knowingly crude here, not dotting all the t's) one file per definition of a linkable entity. This is used as input to the Linker, just naming all the obj's. Similarly, you4 Make scripts would read the manifest for use in dependency chains...

[6] even without going the full way, such a tool would be *very* useful today - hoping to see one built using the LLVM tools.

[7] hint: the trick is in the naming! Let the static be replaced by a non-static and use name-mangling to enforce the visibility rule...

Starlink satellite fails, polluting orbit with debris and falling toward Earth

I missed the Geminids

for all the usual reasons (cloud, a different type of cloud, fog) so maybe this will make up for it when the whole sky comes Alive With Lights.

NIST contemplated pulling the pin on NTP servers after blackout caused atomic clock drift

Re: [System.Globalization.CultureInfo]::InvariantCulture

In Ian M. banks SF novels (you usually need the 'M' to get the SF as opposed to his weirder stuff) the main interstellar culture calls itself "The Culture". Effectively presided over by giant machine minds called, um, "Minds".

(Hey, it could be worse - at least the lead character in any story isn't called "Hiro Protagonist"!)

UK prepares to wave goodbye to 3G telecoms as tri-hard tech retires

Re: End Of An Era?? Really??

> How are you going to stop particular devices?

Same way they do now: my 'phone won't connect to any network I've not paid for. Simply don't sell jew contracts to 'phone holders and if someone does get their hands on a SIM and can manages to make do with basically a low-bandwidth, no voice, limited service with no help line unless they can quote their Big Boys SLA number then good luck to them, they are still paying the contract fee...

> And who is going to pay to maintain the infrastructure when you have no paying customers?

>> meters and other hard-to-upgrade devices

Which pay the same way they have for decades and the same way you are paying now (and probably into the foreseeable future); the majority of the physical plant (metal towers, local backup generation etc) is common to sll the Gs and 2G just bears the cost for sharing its (small) part of that. The software already exists, goes into basic maintenance instead of trying to gain new features.

> In addition, the spectrum has generally be reallocated to other services.

>> you'd need only a tiny sliver of a 2G band available for that stuff, and the rest could be redeployed for 5G.

Tired of sky-high memory prices? Buckle up, we're in this for the long haul

> where is all that compute going to go?

As already replied, dumped.

But it isn't as simple as "why bother trying to flog this stuff cheap to consumers, just bury it - and bury any embarrassing record that we stupid enough to buy into the bubble".

As TFA sorta notes, the DRAM in these AI barns isn't going to plug into your boring old ATX gamer's mainboard. Similarly, the CPUs are weird (ludicrous numbers of cores, which can only really be fully utilised by highly specialised algorithms or by hosting lots and lots and lots of VMs) and need different sockets (and software), the "GPUs" haven't been very "G" for a while now (all compute, no video out circuitry, no HDMI fees paid). And the different chip makers have been chucking out more and more oddities, each different from the last, which they sell en masse to a few barns before moving onto the next iteration - they aren't made to fit into existing mainboard sockets.

Unless someone ends up with a huge stash of identical parts, big enough to be worth the bother, we won't be seeing any consumer (or office worker) grade mainboards being designed to make use of all that silicon.

Not saying I don't believe this, but can you provide references for those of us who missed seeing where the "at less than 25% capacity" was reported.

Cornish recycling drive sows confusion among Reg Standards Bureau

> who would YOU pick as the patron saint for a holiday best known for hedonism, libertinism, decadence and debauchery?

That's one of those sneaky questions, isn't it?

You want us to say (hmpphh) but then the klaxon will go off!

> not past-ies

Who says "past-ies"?

Isn't is "pass-tees" or "par-stees" (or "parrr-stees", the number of 'r's giving you localisation that GPS could only dream of)?

"Past-ies" sounds like you're reminiscing about golden treats from years gone by (or maybe just Ginsters from the Student Union).

Re: we buy cheese in grammes and measure distance in miles, we drank beer in pints.

Which is where the US went wrong.

If they'd only waited a few more years and had the US Revolution *after* the French Revolution, then the French could have brought the metric system over with them and the 13 colonies would have leaped at the chance for another way to distance themselves from the English Oppressors. Which would have then better prepared them to invade the remaining, what, 90% of the continent as they would already have their ammunition properly sorted out (9mm etc).

Never understood "cream first", it is just Bad Engineering: even the yellowest of clotted cream can not support a sufficient mass of jam on top, without giving way. Especially as you can't spread it fully over the scone base without cracking the protective shell. But a fully formed deep plateau of jam first can then be heaped with a delightful peak of cream that reaches to the heavens.

After that, the *real* test begins: can you manoeuvre the massive morsel, without the baked substrate giving way, so that all of the goodness goes into your gob and not have any smeared upon your moustache*?

* a young child or elfin maid may, of course, allow a single blob of cream to remain decorously perched upon the nose, to be wiped off with an aunt's hanky or kissed away by a hopeful paramour respectively.

DOE recruits cloud, chip, and AI giants for Trump's Genesis Mission

Multi-agent virtual scientific collaborator

A "virtual collaborator" - so that is something that, to the outside world, looks just like a collaborator (takes in grant money, demands to have its name on publications) but when you try to look inside it doesn't have any actual substance at all.

OTOH if they let the "agentic tools" have their head, ordering stuff off the interwebs to be installed as dictated by the neatly generated diagrams inspired by scraping the best mathematics images ever published, maybe we will see the DoE complete its transformation of Fermilab. Brought to you by Google DeepMind and displayed on Google Maps, Illinois will finally have a proper Mandelbrot Set.

China turns on a vast experimental network it says is an heir to ARPANET

Re: JANET

UK Freeview

BBC 2, 23:05 Friday 19/12/2025 - Don't dream it, be it! Audience Participation Script & FAQ

Sky Arts, 23:10 Saturday 20/12/2025 - the story from stage show to silver screen

User found two reasons – both of them wrong – to dispute tech support's diagnosis

> Anyone know how to remove Chrome from a Chromebook?

Fire is a beautiful thing...

Re: Idiots are not always wrong

Sometimes, one has to admit that zebras do exist and, yes, that oversize bar code did just eat the office potted plant.

BBC tapped to stop Britain being baffled by AI

Re: Trust

> my trust went when they made Pratchett's "The Watch"

That was BBC America, which is (now) an entirely separate entity from BBC Studios (the commercial arm of The BBC, which goes out and sells studio time to anyone, as well as making programmes for The Beeb) in the UK. Even at the time of that program, the BBC Studios held at most 50.1% of BBC America. The goings-on of BBC America are completely divorced from any discussion of the UK licence fee or the content of The Beeb over here, except for the fact it allows Aunty a prepared outlet for flogging into that market, taking some of the strain off the licence fee. Stuff that BBC America makes/has made for it does tend to get shown here, with mixed results (as everything shown anywhere is going to have mixed results).

How and why "The Watch" ended up so - unexpectedly different - well, there has been plenty of discussion on that (the best result of which has been "shame they didn't change the names then it would have been received better", sort of like how Sanford and Son didn't try to convince anyone that it was really Steptoe and Son) but for the purposes here, BBC Studios just made what the script said.

Re: Why a TV licence?

What authoritarian "limiting of choice"? Gonna have to put that one into simpler words, 'cos my poor befuddled brain can't see where that is happening.

Re: Why a TV licence?

Disconnect the aerial *and* avoid streaming events live. But that is still trivially easy to comply with.

Then they can leach on the BBC's website, learning materials, news reporting all they want.

Re: Why a TV licence?

I agree that those "enforcers" are heavy handed.

But, take that up with Crapita and tell the government to keep them under control.

That work is not undertaken by the BBC.

Re: Why a TV licence?

The children getting Micro:bit? Revision aids? Anything school may decide to use (But I don't have kids! Ok, so long as you never will and none of your family - or friends - ever do)

Ok, yes, you can scrape away and find people who *really* do not ever use anything from the BBC, who don't know anybody who does (don't know anybody they care about whose use of any BBC material is, to them, worth supporting). And yet who still watches live broadcasts and is therefore "snared" by the licence fee. Just be *absolutely* scrupulous in verifying they don't (no listening to BBC radio in the works canteen, no reading the news snippets offered in a web search page, no overhearing anyone who is talking about what they heard on R4 the day before, no second-hand enjoying of any popular culture that came from the Beeb...).

They are, as is everyone who *does* admit to consuming BBC material, then just left with "the common good" (like the submarine captain waiting for the Today programme before letting off his nukes!). Yeah, yeah, damn commies.

Re: Except

> My cultural enrichment is of no consequence to him

Sigh.

Unless someone is quite staggeringly dull, tedious, boring and stultifying, we *all* benefit from being within a culture that is, well, cultured. Even if it is second or third hand experience.

Heck[1], even people watching Gogglebox are unwittingly exposed to the occasional bit of culture, which they've got to remember so they can talk about it later!

[1] warning: incoming middle class snobbery alert! Awooga!

AI-authored code contains worse bugs than software crafted by humans

If these LLMs are meant to "know" how to create a working program, why should we have to still tell it how to do the *most* basic things about the process?

If I am running a compiler[1] I would *really* not be impressed if I had to tell it "ok, try again, but this time remember to do the phase two parsing after running the lexer".

Using an LLM in the expectation that it'll manage the difficult bits (picking an algorithm and a representation, selecting libraries, gluing all this together...) whilst knowing that it isn't capable of the simplest bits (i.e. just bothering with a test run *and* remembering to ingest its results) all on its own. Bit of a strange mismatch there?

And having the "specialised code-writing AI" turn out to be the same LLM, but this time running from a batch file that loops to save you typing in "Wro-ong, do it again"[2] isn't any better.

[1] in the most basic fashion, such as just "cc fred.c -o fred", to avoid getting bogged down in pernickety arguments about how you can actually pass these options on the cli and it'll ... but then you have to ... yourself...

[2] "you can''t have a pudding until you've eaten your meat!"

Re: So I'm not a programmer

> If AI could just get to the place where you could feed it the existing source code as the 'spec' and it would rewrite that in a modern language in a modular/maintainable way...

Yeah, well, we *could* have been spending our time creating a system that could do that sort of thing, but nobody could really be fagged to do it properly. Ok, that is a *bit* rude(!), but way back when (1980s and on a bit) there was much talk of taking in source code and applying lots of lovely semantic analysis, leading to a representation that could be re-arranged in just the way you describe. Even adding in comments.

Trouble is, the demos that could be done were rather trivial, compared to the real-world programs that *need* such attention. And one big, big reason for this was that compilers for the ever-increasingly-complex popular programming languages were moving further and further away from being able to provide even the initial stage, conversion of the sources to a manipulable representation.

Consider the simpler problem, of generating documentation of the form that programs like Doxygen do: it tries to build up a representation of the program and how the parts of it interact but first it has to parse the language used. Oopsie, there isn't a nice compiler we can pull apart to get something that handles the entire language; this is a pain, but you can get a long way towards Doxygen's goals with a rather less rigorous parser, but that leaves far too many holes. Some attempts were made - gcc-xml for example, which could spit out (some of) the AST[1] - but unless they are taken onboard as a goal for the compiler team it becomes a never-ending race to keep the modified compiler up to date; so the project dwindles and dies. One of the great hopes of the LLVM project, in its later manifestation with the release of clang, is to have a compiler that *does* make access to the pre-parsed form a core goal. And LLVM is, indeed, giving us that, as can be seen by the replacement of gcc-xml with CastXML (Note: CastXML is limited in what it does, as it is only maintained by and for one project, ITK, but it shows the way).

You can't really blame the compiler writers, there is enough to do keeping up with the language specs, but just consider how long it has taken us to get to the point where our programmer's editors can do "simple" things, like letting us rename a variable and have all the uses of that variable be updated across all the source files. If things had gone slightly differently, we could have had that as a bog-standard feature of every screen editor since the 1990s or earlier[2] and it would be standard practice to let your compiler be invoked for that purpose[3].

Anyway, cutting the story short (too late!) we don't want LLMs tackling that rewriting task, we want to have a pile of front-end processing, based upon sound compiler theory and practice, doing it. And, as part and parcel of that, make use of techniques from the AI labs to work on the processing: pattern recognisers and rewriters, heuristics that can be reversibly applied... Once you've done that, you could even use Machine Learning to help build up and apply those mechanisms. And, as with extant compilers, you know (bar the inevitable bugs) at each point in the process you *still* have a perfectly correct representation of the original program, so even if the generated modern language output isn't as modular as you'd like it is still good: and you can try again with another set of command options and see if it improves.

But, instead, the best we appear to be able to hope for is that "AI gets to a place", realising that it is far more likely that people will just try to keep shovelling cycles and random text scrapings into an LLM, fingers crossed the few remaining human programmers will be able to fix the glaring holes in the "translated" version of our program (but it *must* be modular now, as it has been broken into 7532 files, each containing no more than two screen's full of text, just like all the old company coding standards scraped from the web said)[4]

[1] a weird consequence of software "getting better" but not if you look at it from another p.o.v.: when we *had* to write multi-pass compilers, we *always* had to have a full representation of the complete AST, complete with line numbers (but lacking the comments), in a defined Intermediate Language format, which could survive being written to disc and read back in.

[2] I have a friend who was working on a "syntax-directed editor" in the mid 1980s who was stymied by the lack of parsers.

[3] Obviously, not the commercial compiler and/or editor authors, they'd shun this or only make it work with their own products <cough>Visual Studio</cough>, which is why this whole thing is a pipe dream.

[4] dang, now I'm implying the LLM can "understand" those scrapings...

Re: So I'm not a programmer

> That's why such dusty deck rewrites fail so often.

Don't worry, if you haven't got a dusty deck ful of random mods made with a pair of compasses, but instead really now understand what the system was meant to do all along, the project will be hit by good old-fashioned Second System Syndrome and will still fail.

Re: I've said it once

Oh boy.

So he was told about the ideal goal of perfectly self-documenting code and promptly grabbed the wrong end of that stick.

At least, one *hopes* it happened that way and he wasn't really taught...

Re: I've said it once

So you recommend his students leave all their code wonderfully legible but incorrect from day one, on the basis that "it'll end up being wrong tomorrow anyway, when the client changes his mind"?

Re: I've said it once

Worse, habitually leaves the comments untouched when they rework the code, so there is no longer any connection between the two. Even just not removing the /* TODO: fill in the body of this auto-generated empty method */ wastes so much time: did they leave it because this *isn't* fully done yet?

Bonus points if they are still using the variable names that were already present, but now for completely different values/purposes.

Attacks pummeling Cisco AsyncOS 0-day since late November

Another day

Another hole in Cisco.

Jassy taps 27-year Amazon veteran to run AGI org, which is now definitely a thing that exists

Quantum AI

*** The Physics Interpretation *** (popular in Copenhagen)

It works perfectly in the lab, all the overnight test runs are successful.

Only problem is, every time we open the door to look at any of the results, the whole thing collapses - and we have to start all over again. It appears that we can know that the AI is running, ready to accept our prompt, *or* we can know where to type the prompt, which terminal is active, but never both at the same time.

*** The English Language Interpretation *** (popular with that Collin guy in Oxford)

The smallest possible amount of genuinely useful functionality we can flog to the public.

*** The Common Speech Interpretation *** (popular in boardrooms and other ill-educated spaces)

A really big and important thing that we don't really know the meaning of but it sounds expensive and sciencey so we must have one, doesn't matter what it costs.

Purdue makes 'AI working competency' a graduation requirement

We at Purdue must lean in and lean forward

Reads like President Mung Chiang is already taking full advantage of AI.

An AI that hasn't quite got the hang of the Hokey-Cokey.

Meanwhile, the advice to the students would appear to be "shake it all about, see what drops out and maybe that'll do to let you graduate". All perfectly clear.

Blockchain company Nomad to repay users under FTC deal after $186M cyberattack

Shouldn't there be an automatic repayment

From every company that ever flogged anything to do "blockchain" and "crypto(currency)"?

We howl about the prices attached to AI and how it'll burst, but at least there are considerable hardware assets present there. Cryptocurrencies are closer kin to the well-known tulips and the instant that faith is lost in them they'll evaporate far more efficiently. Sure, there'll be some mining rigs left, assuming that the loss of faith wasn't caused simply by those rigs (and the piles of storage accumulated for other "proof of" mechanisms) being sold off for "AI" use after the miners realised they could have one big payday on eBay as the next DeepSeek-style AI Hype occurs.

Re: Very selective enforcement by the FTC

Yes, but similar losses to Microsoft would be a trivial amount compared to their total value: i.e. MS get to claim that they really are 99% secure, which is deemed "good enough", especially as MS's core business isn't in "securely" moving assets through their own systems[1]: assets going through *your* systems are sre your responsibility, so if you didn't verify the software running on them, or pay for an SLA relating to security then it is your fault[2]

[1] Of course, with 365 that is what they *want*...

[2] not agreeing with line of argument, but pointing out what the responses can include, which do go a wee bit deeper than just your line of argument

X sues to protect Twitter brand Musk has been trying to kill

But you are only allowed to see the Right Twix; nothing exists on the Left*

* which is very true from the perspective across the Pond, even given how hard our lot are trying.

Or chi, as in Knuth's TeX, which rhymes with the word bleccchhh.

"If you say it properly, your screen should become moist".

Or go the whole hog: "Ecchh (huucchhh spit ding)"

Re: Trademarks

> When companies change their name or merge it's also pretty normal for them to retain ownwership of previous brand names and logos as ...

Those are very good reasons for a company to want ensure that it also retains the old trademarks associated with those names & logos. Not that there is any legal requirement to w.r.t. old contracts, letters etc: unless they were incredibly badly worded, those don't suddenly become invalid just due to change of ownership or mergers. Although it is commonplace for contracts be promptly reissued under the new ownership, that is also when the new owners inform you that the terms have changed (pray they do not change them further)*

But those good reasons for *wanting* do not automatically mean that the new company *does* retain the old trademarks indefinitely: the company has to still make the effort to keep them active as much as previously, and in the public eye. For example, by listing the old trademarks on current products, ads etc.

* I've had subscriptions shut down by the new owners "because we never issue that kind of subscription".

PS

Of course, all these things differ under different legal systems in different countries...

Browser 'privacy' extensions have eye on your AI, log all your chats

Re: I would assume

> https...

Does it count as URL shortening to have a shot at guessing which single video you wanted to refer to?

PS

Mr Morden, not Mr Moden; his - associates - will be informed of your lack of respect in remembering the name.

Low bar, man, low bar.

Stallone couldn't even keep his helmet on.

China's Ink Dragon hides out in European government networks

su VoiceOfTruth

Here we have another article about China/Russia/Genosha ... what about the US/Wakanda doing the same?

CTRL-D

(the answer, of course, being "this is a story about X, drop us a URL in the suggestions box about Y and if if stands up...")