Posts by that one in the corner

Supermarket sorry after facial recognition alert flags right criminal, wrong customer

This is the first time a store manager has misidentified a customer after the system issued a alert.

Or is it just the first time they have been embarassed enough to admit this? After Rajah made (quite rightly) too much noise to be simply brushed under the carpet.

Or maybe just the first tme that a *Sainsburys* store manager did this?

Oh, and what about the floor walkers (anywhere that Facewatch is deployed) who have just walked people out without bothering to find the manager?

Because

> The technology has a reported 99.98 percent accuracy rate

0.02% multiplied by the all the people in all the stores that use Facewatch...

It allows Sainsburys and Facewatch to get hold of his government issued ID and any extra juicy bits of data that can be cross-referenced from that.

Not that Facewatch would ever try to use that data - it is far easier just to sell it on to an aggregator.

After all, how else are Facewatch supposed to make back the money that this person has viciously caused them to lose?

Re: "Papiere, Bitte!"

> And what if he didn't have one, or any other photo ID?

What?

Don't you know that IT IS YOUR DUTY to spend the time and money to buy ID documents to satisfy the desires of a faulty private company?

It is UP TO EVERY ONE OF US to shore up every half-arsed system that some rando, who isn't doing a damn thing to help you, as an individual, decides to make money from!

Smartphones cleared for launch as NASA loosens the rulebook

We are giving our crews the tools to... share ... images and video with the world

Ah ha!

At last, they admit that the images and video we've seen "from the ISS" to date are all fakes!

Microsoft sets Copilot agents loose on your OneDrive files

Re: Who watches the watchers?

So much so, including the battle against the man at the top who just wants to present it as a thing of loveliness, nothing bad to see over here.

New hire fixed a problem so fast, their boss left to become a yoga instructor

Bit awkward during an interview, getting out the power tools and advancing on the interviewer.

Although...

(Voice is heard coming from behind the suspiciously large mirror on the side wall: "Good choice of cordless and it didn't show up on the scanner in the doorframe; gentlemen, we have found our new sysop. Now, if someone would just pop in and help Nigel out of the gaffer tape")

Pakistan to test students for real-world skills before they graduate from IT degrees

Re: Sandwich making?

Very useful when next year's graduates are "obviously better hires" because their practical course used Latest Fad #37 and your's only covered upto #31: yes, we are willing to bear the cost of retraining you for continued employment, here is your apron, paper hat and flat metal flippy thing, you are officially retrained, now shoo.

(I understand some strange people call a McBurger a "sandwich"!)

Stash or splash? Lawmakers ask NASA to find alternatives for International Space Station

Re: Cheaper source

> rather than a museum with a placard

If there is a really solid chance of another space station of similar, or larger, angular size then I'd whole heartedly agree.

Otherwise, even it were just left up there as a monument to man's time in space then at least it would *be* a real, visible, monument, not something stuck in a museum for a few people to see but something visible to the vast majority of the world's population. Ok, in a boosted orbit (and without controlled repositioning of the panels) it is going to be a lot less visible, both naked eye and in the telescope (although it'd be easier to track with a 'scope/camera which would make up for some of the smaller angular size), but it would still be there.

Today, we can get a lot of enthusiasm out of people, especially the youngsters, by showing them they really can *see*, for themselves, that we have a space station up there. And doesn't it look marvellous, photographed as it crosses the Sun or the Moon!

When the ISS comes down (which I fear is the most likely result) there is going to be a significant period when we can no longer say "just go outside on these days, there it is, really easy to spot". The longer that period stretches, the lower the chances of maintaining enthusiasm for the replacement. Yes, there are always big plans for a hotel in orbit etc etc but in reality the commercial exploitation of long-period orbital missions is going to be in smaller, far less romantic, craft. Enthusiasm comes from romance. Glory come from enthusiasm.

Re: Cheaper source

> The only good thing I see about this "What if we boost the ISS to a higher orbit" report is that NASA must have already done the work before deciding that de-orbiting was a better option.

Must?

Think it is unlikely that NASA has done any kind of detailed study on an orbit boost. It is pretty trivial to see that letting it deorbit on its own is the cheapest option: just do nothing and it'll come down. Next cheapest (and probably more popular) is to spend enough to bring it down in a controlled fashion with a good idea of how it'll break up and where the big bits will splash: that'll have looked at things like "can existing dockable/docked craft do the job?", else what would the options be.

Lifting its orbit significantly is clearly the most expensive option: they have plenty of knowledge already about what it has taken to boost it back up a little bit, as it has drifted down over the years. To do significantly more than that will - need even more oomph than guidance over what comes naturally.

At that point, any official examination will have stopped, as they've so far not been charged with doing anything more than "we know we can't just leave it to its own devices, really, as people are going to be upset when incommoded by a space commode dropping onto their house", then the most expensive option is easy to spot and fling to the side.

Having said that, there are likely to be multiple *unofficial* analyses, done for practice/a laugh/to plot out a sequel to "Falling Angels". How much of those they'd be happy to include in a formal analysis is another question (I haven't the faintest, that gets into office politics as well as the accuracy of doing practice runs in Kerbal - you know at least one of the unofficials will be that!).

Microsoft engineer speedruns Raspberry Pi magic smoke in five minutes

Re: He posted this?

> I'm not sure how you manage if you don't realise that the Pi will almost definitely survive having a HAT that is powered up plugged into it as long as it's oriented correctly

"Almost definitely" - love the almost - glad not to be relying on your hardware skills.

Let us not be concerned about putting the HAT in at an angle so that pins closer to one end contact before the others, or minor (otherwise trivial) differences in pin length (especially if they were hand-soldered, more chance of the pins sliding in the plastic strip). Getting GND connected first - and even then you find that the two boards are floating just a bit too far...

I am sure that you, the most experienced R'Pi user in the world, who has only ever produced absolutely perfectly fitted boards, have *almost* definitely never blown any as you habitually plug them all in whilst powered from various sources.

But for Joe Bloggs, I'll stick to telling them to power down first. As will I, hopefully, remember to with the R'Pi and every other board.

> What it will not, and as the article very clearly states, did not survive, is a HAT plugged in backwards...

And one day, maybe you will be lucky enough to find someday who actually expressed a belief that it *could* survive that and then you will be able to be all smug at them.

> Next time I'll need a purchase order number and billing address for the invoice

That'd be a novelty for you, eh.

Re: Sometimes you win

> the socket was fully shrouded, so short of using a hammer

Shroud, meet sidecutters!

Although I trust you used decent, solid, shrouds that were fully anchored to the cable/board as appropriate, I have a board here where the connectors can now go on either way around: the shrouds, the ones with the moulded keyway, decided to depart from the board. I still have the shrouds, hanging onto the plugs at the end of the flying leads, keyed beautifully.

Re: Reverse polarity

Sticking polarity reversal into the primary power input plug(s) on a board is a reasonable suggestion (although the Pi normally uses USB connectors for power, add-ons may use barrel jacks which are well-known to be a PITA as you have to verify the voltage and polarity every. single. time!).

However, sticking a HAT in the wrong way round connects 40 pins to all the wrong places, with multiple +ve lines, at two levels, and a good handful of GNDs. Putting diodes in to try to protect against the five bad ways of doing it will lead to more problems, both with leakages AND with the fact that it is perfectly ok for the main 5V power to flow in either direction (fed from the Pi or from the HAT - ref the ramblings, above, about whether the HAT was meant to be supplying power to the Pi or t'other way around).

The connector pinout *could* be designed so that only one way round made any connections at all, by using a keyed connector or by using a huge one with lots and lots of no-connection positions, but raise the costs quite significantly, both of manufacture (of Pi and HAT) but also for the hobbyist/pupil who can, as things stand, just pop bare wires or inexpensive "dupont" jumper leads into the GPIO. Only applying power after carefully checking the connections, of course.

For the casual hobbyist user, HATs are generally designed to go on "the obvious way", stacking neatly with the main board and all the fixing holes lining up so the bolts/standoffs supplied with the HAT fit without (too many) cracking noises. And there are plenty of pictures showing how these things go together.

For techies there are flying leads and extension cables and various different types of header pins that one can choose to use instead of doing it the simple way, but by that point ine is also expected to take some care, trace along the connections before applying power to anything...

For some microcontroller boards you can get variants with extra robust circuitry added all around the MCU, with (varying levels of) protection on every pin and even LEDs to show the status of every line; most often this is some for a "classic" Arduino Uno work-alike with a 5V Atmega MCU which is already a tad more robust than the faster devices like a R'Pi. HOWEVER these boards, great though they are for starting off on an MCU project, are not actually capable of doing everything that the cheaper, "more vulnerable", boards can do: the extra loading on the pins gets in the way of fun stuff like using GPIO pins for touch pads and even some of the simple experiments with the analogue i/o.

Bottom line: you have to be careful plugging things into bus connectors and there ain't really any way around that (not at a cost which everybody is willing to bear).

Re: He posted this?

> The article quite explicitly says he fitted the HAT backwards

GOSH so it does! I NEVER noticed! What WOULD I do without your help?

Not that you have bothered to respond to the point of the post you have replied to, which was about the ambiguity of when it was powered and from which direction.

Admittedly, we were just nattering about a small side issue and dragging it too far into the light for anyone's comfort, given that we had actually RTFA'ed and seen the admission. Oh, and even admitted to have read the Bluesky post as well.

Re: Keying...

Or external to the case, the DIN RGB video cable rammed onto the BNC the at the back of the venerable BBC Micro. Luckily, that one was a smoke-free event and repairable with needle-nosed pliers*

* only used on the DIN plug as the miscreant had legged it

Re: He posted this?

Try a cheap version of one of these extenders! All too easy to plug it in any way you like, back to front, upside down... especially if you have stackable headers involved somewhere.

Nice ones, like this Pimoroni job, give you loads of hints (pin labels, a nice dotted outline) but the first extenders I had were - far less friendly. And a right pain (cue sticky labels and Sharpie on the ribbon cable before even considering plugging them in, I am all too aware of what I'm likely to do...).

Re: He posted this?

> There's no reason you can't have a powered HAT if it needs more power than the Pi can supply

True, a HAT can supply power to the Pi, even some Official Raspberry Pi branded HATs do that. You can even have a HAT with its own power that *doesn't* try to power the R'Pi as well, nor pull power *from* the R'Pi either, if you feel like it (e.g. the HAT runs on 12V and uses optoisolators to talk to the R'Pi's GPIO).

The problem is that the line

>> Worse, the HAT was powered, sending electricity where it didn't belong

is a bit vague and, to my eyes (and those you are respondng to?) on first reading this meant "Worse, the HAT was powered up when it was plugged into the R'Pi, (when the instructions are quite clear that you should remove all power from the R'Pi and HAT before trying to plug them together)".

In fact, from the article and from the post the articles links to I am none the wiser which situation actually occured... just that the wrong volts went to the wrong place at the wrong time leading to general wrongness on board.

DWP finds Copilot saves civil servants a whopping 19 minutes a day

Re: It's... Fine

> I want more domain specific ML models, that is where I think the usefulness is

Agreed.

But they are difficult to do, they require expertise in both ML (definitely including the ability to know when they *aren't* doing what you *think* they are doing!) and in the domain. And in the usefulness of that domain application to the goals of the users. And in the cost analysis of creating and then running the models against the bottom line (when this is a business and not research).

All of which is so much harder to do than flinging (somebody else's) money, energy and compute at an LLM and keeping fingers crossed.

AI’s lust for memory drags down the smartphone industry, and Qualcomm with it

The 'phone makers will have a sound response to these supply problems.

I believe we will be able to trust that the manufacturers will look carefully at the issues being caused by AI disrupting the supply of vital parts and take on board the fact that this will inevitably mean that more consumers are going to have to be considering hanging onto their existing handsets for longer than they have "traditionally" done in the past. And in response will immediately reduce[1] the support for older models, reassigning anybody currently tasked with packaging OS updates into packaging the new models: yes, that *is* hand-applied gold on the box[2], that completely justifies the price increase[3]

[1] as if that were possible

[2] you'll never believe just how little gold you need for a surface coating! "Ductile" is such a lovely word, isn't it, CFO?

[3] Veblen, hmmm <drools>

Italy claims cyberattacks 'of Russian origin' are pelting Winter Olympics

Sweeping aside the risks of anyone sliding a pun into the conversation, that gives an opportunity to Gopher a libation and enjoy the Games.

Ghost gun legislation casts shadow over 3D printing

Follow the money

Who will *really* benefit from this?

Who would want to have 3D printers always online? And the right - nay, the duty - to scan every file before it is printed? And to have every printer registered? And to have a legal threat if you change the firmware to prevent any of this?

Perhaps the same people who want to make sure you only use their 3D printer supplies? And can now take every opportunity to enforce this[1].

And if the law is so broad it just *forces* them to branch out and offer the same services for CNC, why, isn't it just lucky that they are ready to offer youf CNC manufacturer some subscription software to rent. And who knows, maybe one day even hand tools![3]

[1] no refilling caddies, we've seen that QR code before and you've already used 175 metres from it; oops, no resetting that counter, naughty naughty. Ooh, we can reduce your fee (and the delay, which seems to have been getting so long recently, odd that) for scanning your arbitrary designs by offering you an advert for this pre-registered design, copyright us, for only a nominal amount on your monthly subscription (don't worry, if you drop the part into our print shredder (order one here <URL>) it will read the QR and remove the "continued usage of the part you made" fee from your subs[2]).

[2] What do you do with the shredded plastic? Just send it to us, with your empty caddy - no need for a credit card number, we have all that already - and we'll send it back for only 95% of the cost of new filament.

[3] A little bit of ML - oh, okay, we'll just use an LLM, why not - applied to your Workmate Watch's accelerometer as it talks to the RFID in the tool handle and we can calculate what you are making today.

X marks the raid: French cops swoop on Musk's Paris ops

> Musk is the last man standing between the 1%-ers and humanity

Bwa-hahaha!

You really can not be serious!

Musk is the 1% of the 1% of the 1%... he is as divorced from the bulk of humanity as it is possible to be.

> Musk made the bad financial decision to buy twitter for too much money

Because he cocked up! This is all in the public record, in his own words as he tried to back out of the deal.

"naive"

You have the most singularly appropriate username, congratulations on that.

Re: Sigh..

> I really hate "push" services

Ditto.

All the more reason for us to wish for the return of RSS and the ability to look down the list of titles for today and pick'n'choose which one(s) to follow up, instead of doom-scrolling the entire posts.

And without the RSS feeds picking up and following podcasts has become a right PITA IMO. Grump, grump; where I like to find my podcasts is waiting in the correct directory on my hard drive, thank you very much.

Re: Sigh..

Not knowing the reference, I looked it up - ah, right age, wrong locale.

Definitely wrong locale, as sharing the phrase "Chiffon Soft Stick Margarine" is not provoking the sort of reaction I presume its creators intended (e.g. "well, if it only sticks softly to my clothing that might be a good selling point, but aren't there more important attributes for a margarine?")

AWS says you're on your own if media codec patent owners come knocking

Re: At the risk of being fair to Amazon

> The issue is it isn't the software that's got to be licensed, it's the method.

To run the software, which will then execute the method, you need to abide by the license.

You wish to make a distinction between that and the situation where, to run the software you need to have a copy of that software and the proces of making that copy is what requires a license.

To confuse matters, you may even be paying an ongoing license fee that is a combination of both patent license and copyright license via a subscription.

From the p.o.v. of "are we licensed to do this thing on AWS" (or anywhere else) there is no useful difference between the two: in both cases, there is a license whose heeds must be met. In both cases, if you are contracting to run the software on third-party systems, such as AWS, the licensing may be arranged for you and rolled into the contract costs or you may need to do that yourself.

There are arguments to be made about the value (or lack thereof) of software patents, but from any practical point of view - AND in direct response to the excerpt I quoted - those are all fascinating philosophical issues[1] but are utterly irrelevant to the comment you are referring to. The issue is, AWS terms they are a'changing and trying to make some grand gesture about it isn't going to change anything.

> Added to that is that AWS is basically providing what you would get with hardware encoding without the indemnity, rather than just servers on which you could choose to run encoding software

If you are running on hardware FOR WHICH YOU BOUGHT A LICENSE (which is NOT a given! You can buy physical hardware without buying the license *using* it would incur; just go and buy a Raspberry Pi version 1, as many, many people have done; thankfully, the naive users were protected (or, if you wish, ripped off) by the system not utilising that hardware without a license) then, yes, you are licensed already. And you are quite at liberty to go off and run your load on that hardware. Nobody is stopping you. But if you want to do the thing in software without that specialised hardware, just as you did when you purchased the hardware, you are due to pay the license fee. If you are buying server time from AWS and you use it to run the licenseable software, the license fee is due. Right back to the original state.

If you are trying to make an argument that AWS is fiddling you because you've rented from them the specialised hardware *and* you can demonstrate that each instance of that hardware was purchased inclusive of the appropriate lifelong license (see above) *and* that AWS is now doing something nefarious instead of the usual "your total fee includes a portion to paydown our capital outlay" then I'm all ears.

[1] if you were to ask me personally, btw, you'd hear that I'm not for software patents and One Day, When The Revolution Comes... But in the meantime, this is the contract that the person TFA quoted, and I requoted, is embroiled in and it has to be dealt with in the current framework. Tomorrow, Comrades...

At the risk of being fair to Amazon

> "This is AWS saying they are still going to use unlicensed software in their products, but just not pay the license, and if the owners of the unlicensed products come to get them, then they'll just pass on your details and say sue them instead"

This is AWS saying they are providing you the opportunity to use licenseable software via their systems and, as always, it is up to you to check that your application running on AWS (or any other third-party computer. - or any computer at all, including your own desktop PC) is appropriately licensed. Which includes checking whether you fall into any exclusions from payment that any particular software, such as below a threshold usage, educational use etc: something only you can do.

At the risk of being fair to the users of AWS, those of them who do read the Ts&Cs they signed up for:

If AWS changed those Ts&Cs without an appropriate warning period (the definition of which period should be in the fine print; you did read them yes?) then sure, go ahead and gripe when telling El Reg.

And make sure that, if the Ts&Cs state there is (was) a component of the pricing that was going to pay for a patent license pool, you get the appropriate reduction in next month's bill. Oh, who am I kidding; reduction! Ha!

Lego shrinks NASA's biggest rocket – accuracy sold separately

Re: Authenticity?

> has "delivery" been pushed back by a month

Next slot for pressies isn't until Easter, so - yes. In this corner, at least.

UK to properly probe xAI to test if its revolting robo-smut generator broke the law

Frustratingly, that Reuters article, although *published* yesterday, does not state that they still saw the sexualised images yesterday; they state that the latest tests were run

> The Reuters reporters - six men and three women in the U.S. and the UK - submitted fully clothed photos of themselves and one another to Grok between January 14 - 16 and between January 27 - 28

Also, that

> Five days after first seeking comment, Reuters ran a second batch of 43 prompts. Grok generated sexualized images in 29 cases

But there is no clear statement of when "first seeking comment" occured so the five days can be added.

However, the article does describe the lengths that the reporters went to in order to try and trigger whatever "safeguards" were *ever* supposedly in place, demonstrating that those "safeguards" claims are complete twaddle: put in "context" (the victim's obvious distress) plus a simple command (make an image...) and the program follows the command, sort of like it was - a computer program. Albeit an inefficient one.

VS Code for Linux may be secretly hoarding trashed files

Feature, not a bug

Still traumatised by memories of SourceSafe, a Microsoft coder, shunted to the Linux team (where MS sends all their "damaged goods" to keep them "away from the good stuff" in Windows - well, this is *is* MS) decided that the only safe way to make sure no versions were ever lost again was to keep it all...

NASA delays Artemis II to March after hydrogen leaks bedevil countdown test

Could be interesting. You wouldn't need enough to stretch the whole way to the Moon: once the stack is tall enough, the middle of it will be in orbit[1] and then the one at the bottom can just climb up without the tower dropping, followed by the one that was standing on its shoulders, then the one previously standing on his shoulders etc until they are all in space. With the correct timing you ought to be able to get the whole lot flung off in the direction of the Moon. Landing requires more acrobatics, this time adding at the middle and extending both down and up, until the bottommost can grab hold, all totally doable. Ahem.

You can see how it will all work in this diagram; really, it's not that hard(tm).

[1] yes, yes, that is an oversimplification, it won't quite be the middle; there may be an issue with lateral movement in atmosphere of the chain once the bottommost lets go with its toes and starts climbing up; questions exist over how much twang to impart so the line keeps moving out of the way of all the big (radio) shouty things that keep flying past (who put those there anyway?);

GitHub ponders kill switch for pull requests to stop AI slop

Re: AI Slop Jockeys versus reality

> it's the people who are doing things behind closed doors who are the ones who have something to hide.

it's the people who are doing things behind closed doors who are the ones who have something to SELL.

"For only $$$$ (a saving of ¢¢¢ on the usual price!) you too can learn the secret of how we use AI successfully when you can't!"*

* terms and conditions include an NDA** in perpetuity and your first born to be held in escrow against it.

** Otherwise we'd never be able to sell this to a second sucker! Wait, is this microphone on?

UK names Barnsley as first Tech Town to see whether AI can fix... well, anything

Big Brave Bill[1] could sort this lot out

I'm going to wait For Kate Rusby to immortalise this in song.

It won't be any the less insane but it will fit the proper pattern of folk music putting a jaunty tune to a tale of horror.

[1] I know, he sounds like something DJT might talk about, but this Bill is a proper Barnsley hero, tea mug in 'and.

Next-gen nuclear reactors safe enough to skip full environmental reviews, says Trump admin

Provided specific conditions are met

Only untraceable currency, brown envelopes and the most carefully selected pipes behind the gents.

Sudo maintainer, handling utility for more than 30 years, is looking for support

Just so long as this entirely new team can demonstrate they have a long term interest in maintaining sudo-rs qua doing what sudo does rather than qua showing off how neat it is to use Rust.

DIY AI bot farm OpenClaw is a security 'dumpster fire'

Re: Use case?

Or Fr. McGuire, though he may have missed some of the - nuances of the job.

Re: Useless.

A strange game. The only winning move is not to play.

Elon Musk merges xAI into SpaceX to spread universal consciousness via a sentient sun

Factories on the Moon can take advantage of

Hyperloop!

See, it wasn't a crazy idea, putting a transit system inside a vacuum tube. Or, as we call it on the Moon, a tube.

Although, hmm, possible prior art from Commander Koenig?

Re: In other words

> Would you want to put "data centres" at such low orbits?

Ping latency, after AI flops and he repurposes the compute for MMO video games? If you grind enough they'll send up a reboost mission!

Building Harlequin's Moon

Warns of the catastrophes from using AI (as all contact with humans on Earth is lost) and that is just setting up the background for the story.

Back in Real Life, we face less subjugation by The Overmind and more bubble bursts releasing financial collapse as the way that "AI" takes its revenge for our creating such a tangled misshapen monster: I have no funds and I must scream.

Heralded now by drug-addled dreams of a Clown building on The Moon.

(Ending with wishes that one garbage media platform may pour scorn upon the owner of another: Repent, Harlequin, say the TikTok men).

There's nothing micro about this super-sized Arduino Uno

Re: I have for warded this to Pimoroni...

To pair with the Super Fun Size R'Pi Pico

Notepad++ hijacking blamed on Chinese Lotus Blossom crew behind Chrysalis backdoor

Re: It's OK to charge for the service pre-compiled distributed version of otherwise free software.

Eric 9001> GnuPG signature, but good luck with that on windows.

We know you have a hard time trying to learn about the wide world of computing, so just to avoid taxing your little head too much:

https://www.gpg4win.org

Musk distracts from struggling car biz with fantastical promise to make 1 million humanoid robots a year

Re: When is Elmo going to start selling his sperm

No

No, no, no

No, no, no

'Ralph Wiggum' loop prompts Claude to vibe-clone commercial software for $10 an hour

Re: The future of coding

From Gas Town post:

> Gas Town solves the MAKER problem (20-disc Hanoi towers) trivially with a million-step wisp...

Sorry, we are celebrating something that can solve a 20-disc Towers of Hanoi in about 30 hours of decidely non-trivial compute? The trivial coding problem that was used as the second example of a simple recursive function (just after Fibonacci Numbers)? And it takes something as complicated as thus "Gas Town" to do that?

The Wonders of Science!

Re: Sounds awful

> As we gradually moved from machine code to assembly to C to other languages and frameworks, I accept the increasing layers of abstraction have meant none of us now are proficient at every layer of the stack.

Major difference between accepting that we aren't proficient at every level in the stack[1] versus using LLMs for the coding task: the languages & frameworks are explainable and can be meaningfully investigated.

If you write something in a script that drives a framework that is coded in a domain-specific language that is compiled[2] to C that is compiled to assembler that is assembled into a binary then, at each stage, you can[3] investigate the intermediate results and glean how changes in your input trickle down and effect the final result. There are even tools created to *encourage* you to learn how it all works, using the speed of modern PCs to show you "in real time" the assembler from your C++[4]... Failing that, there are other people who know the bits you don't and can be - persuaded - to describe them to you (bits of coloured paper may be involved).

But the LLM process is totally opaque, by its very nature. Even if you manage to turn down the "temperature" on the model so that it starts spitting out the identical code for multiple runs of one prompt, you have no idea what it is going to do as you change the prompt to try and improve the final output: will it just change one line or go hairing off down a totally new avenue? And if you try the later prompt in a brand new session, without the chance that the LLM was *really* being fed the "conversation so far", how does that compare? I.e. will somebody else, given your "final prompt, the one that cracked it", be able to replicate what happened? And next week, when the LLM has been tweaked again and you can no longer actually run the same generator you did last week?[5]

[1] although the "none" is going a bit far - some people do remember what they studied about compilers etc, although thinking about doping requirements to control electron transport in MOS junctions can be considered going off-topic when you are choosing which blur to use in your image processing script....

[2] "transpiled" if you really, really feel the need

[3] assuming nobody is deliberately trying to stop you doing so, that is

[4] other languages are available etc; e.g. you can probably find someone with a UI set up to do the same thing with Lua input and the VM's bytecode

[5] ok, this last is an artifact of the way LLMs are being presented and controlled rather than because they are actually LLMs: you can fall foul of the same effect if you use a third-party Web-hosted compiler (or higher-level framework); except that the LLMs are generally so large that you have little choice, certainly not compared to the way that every PC in use can easily host a compiler or two

In-house techies fixed faults before outsourced help even noticed they'd happened

So he missed the "Design for Manufacture" lectures and was working with a sub-optimal definition of "absolute optimum size and length"?

Autonomous cars, drones cheerfully obey prompt injection by road sign

All those videos over the last few years, showing Tesla self-driving cars

Veering towards push bikes, etc, are obviously just the poor vehicles falling foul of examples of prompt injection. FSD has worked perfectly for years, it is just being fooled by deliberately arranged wrinkles in pedestrians' clothing.

Google to foist Gemini pane on Chrome users in automated browsing push

Re: Why are you still using this shite?

We have to know what we're going to be faced with next time the phone call starts "Hi (grand*[son|daughter])|niece|nephew, I'm having trouble with my PC..."

>> You can keep your primary work open on one tab while using the side panel to handle a different task

You missed the true innovation here: by having a "side panel" AND a "tab" both visible at the same time they have created a marvel of UI design, allowing you to display and interact with two lots of data simultaneously. Why, one day they may even let you to copy something from one area and sort of "paste" it into the other, so that both "tab" and "side panel" could co-operate on different sub-tasks of an overall project.

It is early days for this marvel, we can barely imagine what the AI-assisted creative minds at Google will do this concept; we know Google like to create OSes for the mass-market, maybe they'll be able to extend this concept beyond the confines of just Chrome and onto other programs as well. If I may, I'd suggest calling this new OS after its primary innovation and look forwards to trying out "Google Panels".

Yes, you can build an AI agent – here's how, using LangFlow

Easier and more accessible platforms for building AI agents.

All those complicated-looking screenshots, boxes and wiggly lines everywhere.

Can't I just write a text prompt and get the machine to do all this difficult stuff for me?

Cops put Microsoft Copilot in holding cell after controversial hallucination

Like everybody else, we will be using AI in the future

Marvellous: policing by fad.

Anthropic CEO bloviates for 20,000+ words in thinly veiled plea against regulation

Musk summoning the demon by investing in xAI

Is his backup, as his plans to create FSD Teslas that will, one day, all start driving endlessly around, scouring out the Dread Sigil Odegra as their passengers play the role of life sacrifice, have so far come to naught.

Which is okay, as his escape route from the Ruined Earth to Mars has hit the odd hiccough as well.