Bollocks
Nice try but it does not wash. LLMs need compute only. The network is already way faster than any LLM can worry.
Scraping for search is already built in so what could a LLM need beyond that apart from being a bit more needy?
255 publicly visible posts • joined 29 Oct 2023
The ESBA and WSBAs are both part of the UK as in just as UK as Somerset or Perthshire. Then you witter on about "the Cypriots". Cyprus itself is a game of at least seven halves: Greek and Turkish, south and north oh and the UK. ... and the rest of the shady factions that skulk around.
In the middle of Nicosia there is (or used to be) a statue to the great (Greek) Cypriot nationalist "Grivas". He is or was shown throwing a hand grenade ... into a bus full of women and kids. Well, you draw your own conclusions on all that stuff. My mum did rather a lot of investigations into it all whilst we were stationed in Cyprus in the mid eighties. Surprisingly not, its all a bit of a nasty muddle. EOKA seem to have been nearly as unpleasant to the locals as they were to the UK.
Anyway, Mr Furious .. fuck off. Its unlikely you have any idea what you are on about.
OK it works for you and you have it correctly configured by the sound if it.
However the things that connect to it are probably not air gapped or immune from the baddies. Once one of those gets into contact with your pariah box, the game is likely over. Your Win 2003 box will have absolutely no AV or any other protection more impressive than speaking a variety of SMB that everyone else has largely deprecated. It will be able to talk NetBEUI with a dodgy accent but that's no use these days.
Mind you I do have a customer with a very expensive machine that has a DOS 6 controller and I put a Samba daemon with all the controls switched off on one side and turning the machine share into a SMB 3.12 share.
Anyway, I'm sure you'll be fine.
Fortinets allow you to restrict logins to the webby console from particular IPs, which even a Draytek will do too. They have MFA available and cloudy keys too.
They are generally rather expensive devices and you do have to hate yourself and security to get it wrong or not give a shit (which is most likely). They are not particularly "opinionated" about admin access security but not too bad. I suggest they get opinionated.
They are not the most intuitive efforts either so if you get one going, you will have probably read some docs and so I suggest you have few excuses. If you can get an inbound NAT working to an internal service then you have read some docs. On the other hand they are very odd (but not as odd as a Watchguard) so perhaps we can allow that the usual closed source wankery, trying to differentiate your offering is causing the usual snags. Obviously Fortis are quite happy to run on a Linux kernel whilst being downright weird (they look more like FreeBSD at the commandline). The riff with Juniper and Dell OS9 here ...
If you are pwned due to not bothering with things like decent passwords or MFA then soz, lol.
I actually run a company and have staff etc to do that but I suggest you contact HMRC's advice line/bot to see what is possible.
Ask it/them whether you can claim for your own time doing your own accounts. To make this possibly work, you will need to do the work out of your regular employment hours. I suggest you could allocate say two hours per week on the weekend. Set a reasonable rate for the work: at least minimum wage but not excessive. You are probably not an accountant so charging yourself at an accountant's rate would be daft and considered excessive. £20ph seems a good start. Make sure you do actually stick to doing the work at the stipulated time or at least you are "in work" at that time - so your laptop should be running and logged in even if for a couple of weeks you are watching iPlayer!
The work is recording expenses/transactions, maintaining records and preparing data and recording it for online taxation requirements imposed by HMRC. That's a start - you'll come up with something better.
The above assumes that you do self assessment and that the flat £1000 expenses a year thing is not suitable for you. It might be best to hire an accountant instead but bear in mind you can only claim back the taxed portion of their fees.
eg You earn £10,000 over the untaxed allowance - that £10,000 will be taxed at 20% (1) ie £2,000. You hire an accountant who charges you £1,000 and saves you £500.
So the tax you pay is: (10,000 - 1000) x 20% = £1,800. The accountant saved you £500 by being an expert so it looks like a bargain: you paid £200 less to HMRC because you hired an accountant. You saved £500 via their advice. So that's a saving of £700 - lovely!
However, you paid the accountant £1,000, so you are £700 up and £1000 down = £300 down.
I've made a few assumptions here but a few minutes with a spreadsheet and a close reading of (1) will soon find where the point of hiring an accountant kicks in and how much you will want to pay for them.
(1) https://www.gov.uk/guidance/rates-and-thresholds-for-employers-2025-to-2026
"The software is a legitimate business expense."
There are two ways of doing expenses: You can simply say "its £1000" no questions asked and move on; or you record all expenses, keep receipts etc. Depending on your circumstances the £1000 can be a great choice because it requires no record keeping and time spent faffing with paperwork.
That £1000 is now going to be about £800 unless the rules for expenses change.
"The team recruited people with prior lucid-dreaming experience because they are better able to control dream content and search for insight while asleep"
Well, that's complete and utter bollocks for starters.
We (probably all organisms more complicated than an amoeba and perhaps even so) all dream in some way - I think that is a given. As to what is a human "lucid-dreamer", let alone why they are able to "control dream content" whatever that means is open to some interpretation.
"Even without lucidity, one dreamer asked a dream character for help solving the puzzle we were cueing,"
The team are taking drugs.
Steve S is a seriously clever bloke. I only know this because I have an AD domain with a kerberos over https KDC lying around doing its thing after finding a blog post he made. I think he is a mostly software geezer but he is a proper nerd and obviously he insisted on diving in on the hardware side by putting his tongue across 5.5V PD. I'm sure we have all done that at some time.
I doubt that anyone here has not looked at a 9V cell - the ones with both terminals on top, and thought ... hmm I wonder what happens when I touch my tongue to those metal things.
You probably don't do enterprise IT and neither does MS (but they are learning, gradually)
You need the likes of sysmon for system profiling and clearly MS have finally noticed that their shoddy MVP called Endpoint Management is a bit lack-lustre and are finally adding it in.
This is nothing to do with screenshots and that. It is all to do with being able to winkle out nasties.
Not to mention your arse! You feel acceleration through your entire body. There is a good reason for the phrase "flying by the seat of your pants" - it still works for ground vehicles.
Although your eyes only scan a small area at a time they do it quickly and you can move your head. At night, when you are going to be nearly blinded by an on coming vehicle's headlights, you know to drop your gaze and watch the edge of the road.
You mention your inner ear, so that's acceleration again. However you should be using the rest of your ears too. I often look left and listen right when turning right (in the UK, I do wind down the window). I use my ears to get a vague idea what is going on and then use my eyes for the final go/no go.
"according to analyst biz Gartner."
Who can stuff it up their arse. 1% of GDP on LLMs? Nope. You can have a Navy or a load of gas guzzling data centres stuffed full of hardware that hallucinates!
If Gartner is all that is left to pathetically beg people to subscribe then the bubble's burst is soon upon us.
"teach kid to place card within range of a video camera"
Putting a floppy into a drive is a simple tactile process, with a satisfying feedback (*clunk*, whirr etc) that is easy to understand and so is the eject mechanism. The labels are easy to follow.
QR codes are a confusing mess, if you don't know what they mean and you have to use just hand/eye co-ord. and aim an invisible "eye" at it. I suspect the child will aim the eye at the picture, which is the obvious thing to do. You'll really need to turn the camera into a proper handheld scanner to have any real chance of getting that to work. You could put the QR code as an overlay on the picture but it will look pretty rubbish.
The floppy disc effort is a fabulous solution and is also generally more accessible than scanning a QR code. For starters, it will work for a huge range of dexterity abilities and will easily lend itself to braille and working well for the blind.
One of my customer's machines (a bridge mill, I think its called) boots DOS 6.2.
I have a Samba daemon acting as a go between, so that Win 11 can even see its input data share. It's NetBIOS on one side and SMB 3.2 on the other. I had to turn off a couple of safety catches in Samba but it is doable and it can fix up filenames and case on the fly.
"If only they'd just added a couple more bytes to the address and left it with room to add more as needed."
Oh do show me how your new protocol will fit within upper layers. Bear in mind your new addressing scheme will need to fit in an ethernet frame and other frames.
Now tell me about processing efficiency: how are you aligning your addresses and other data to make the best use of registers and the rest? How will this work within existing hardware?
It isn't quite that easy as adding a little bit more bigger numbers.
"The problem isn't the lack of compute, it's the lack of science. We just don't know how to build an AGI."
Very true but that does not win arguments in the muggle world that will insist on "well just nerd harder". This argument demonstrates how the current efforts will never achieve AGI because laws of physics, no matter how much cash you throw at it.
Trump might want to restore basic science funding to give the nerd harder thing a chance.
Let's drop the AI moniker (divisive and bollocks) and look at "LLMs" instead.
Please bear with me because I'm going to whitter on as an engineer and I think you are more humanities focused.
I use them as a tool, just as I do a hack saw or a slide rule (I have two) or some of my "make it smaller, deeper or more broken" devices: percussion tools - fencing maul, sledge hammer, lump hammer ... you get the idea.
I bought a second hand Nvidia A2000 with 16GB of RAM and popped it into a box at work (Dell server) that generally acts as a fancy NAS for customer backups over night, so its bored during the day.
With llama.cpp I can run a small LLM - 20B parameters or so locally. It is quite surprising how much general knowledge even a small model can have. I'm actually interested mainly in programming but I do get them to do english to latin and vv or eng to german and vv. I've also asked it to explain physics ("tell me about the bernouiilllii equations" - with deliberate mucking about with spelling) and get a reasonable answer. Questions about small towns in Somerset get reasonable but rather generic answers.
Bear in mind this is in a dataset that is around 16GB in size or around three DVDs - ie a pretty big encyclopedia that works quite fast and can sort of reason too.
ChatGPT, Claude and co have much bigger data sets and their models run to something like 100s of billions and even trillions. They also have a lot of other machinery tacked on too. At that scale of data, you might question the quality and even the provenance of the data inputs. Let's put it this way - they ain't 100% encyclopedia Britannica. That said, neither am I.
So, you can rail against the machine or not. Your last para did rather anthropomorphisise (how the blazes does anyone spell Greeklish correctly!) the beasties. For me they are a handy tool but they do need some care to use effectively.
My first car was a Ford Fiesta MK 1 (UK). It was tiny but there was nearly as much space around the engine as my current EV has under the bonnet.
In the end there was nearly as much wire mesh and body putty as sheet steel and there was a nasty draught up your right trouser leg. Its second engine costed £100 including fitting in a tiny garage near Plymouth Central. It finally passed away just off J30 of the M5 (Exeter Services). It was sounding a bit rough whizzing up the A38 and it turned out I had basically blown two cylinders out of four. Well I got a good year out of that engine. The AA recovered it for me back to Plymouth and I got £25 from the scrappers for it in 1994.
Your parent quite literally gave an example.
Arabic as a whole is largely written R2L You note that numbers are written L2R in Arabic.
Mr Harston notes that quoting foreign phrases within Arabic is written according to the preferred direction of the quoted language. English is a habitual L2R language, so english quotes will be L2R.
Not for me it isn't, says (nearly) ex VMware consultant with roughly 20 years experience.
Over those years I've been on the receiving end of so many MVP efforts that I basically had enough and bailed when Broadcom.
Windows vcentres being required and then deprecated, vcentre being a shit appliance, then the transition from Flash to the current effort. Ooh lets run two Tomcats - one to monitor and manage the other one, which is running on the same system, rather than doing a proper job in the first place. vcentres have 15 odd volumes nowadays - WTF! They are monstrous efforts.
OK so vcentre is wank, buggy and slow as fuck. But it isn't even a proper orchestration jobbie - you'll need another VM for that. Again, another sodding monster and you'll want to cluster that for scale. Oooh what about SDN? That will be twenty trillion squid, and yet more VMs and more shite to manage.
I can't be arsed with it anymore.
I'm going all in on KVM and I sort out the orchestration and management myself using off the peg open source stuff. I always go for offerings where there are two options to avoid enshitification.
" still love that, in 2025, explorer just freezes/hangs if you have a network path that's unavailable ANYWHERE on your drive list."
If it helps, it seems that Dolphin does too (Linux box with KDE n that).
Obviously the notion of a "drive" is Windows wankery - one mounts one's remote file systems. I have to live in a Windows world at work - I use pam mount to connect up a series of shares under a single directory. If my VPN is down and I access that folder, things lock up for a while. To be honest it is actually quite quick to recover and does seem to just work most of the time.
The real fix is SMB over https (oh yes!) There is a Kerberos proxy over https too and I have deployed it at work and it does just work for both Windows and Linux. One day I'll look into SMB over https.
Refs wrt Kerberos proxy:
https://syfuhs.net/kdc-proxy-for-remote-access - MS Blog on KDC Proxy
https://stackoverflow.com/questions/779228/the-parameter-is-incorrect-error-using-netsh-http-add-sslcert - SO discussion about http add sslcert
https://www.mit.edu/~kerberos/krb5-1.21/doc/mitK5features.html - MIT Kerberos v1.21 features
https://k5wiki.kerberos.org/wiki/Projects/KDC_Discovery - KDC Discovery via DNS
https://www.mit.edu/~kerberos/krb5-1.21/doc/admin/https.html - MIT Kerberos documentation on using a proxy
"Actually, LLMs are completely deterministic"
No they are not. I have a box at work that is used for backups - it mostly works at night. I slapped a second hand Nvidia GPU in it and a slack handful of llama.cpp, ollama, MCPs and a couple of webby front ends.
There are a lot more parameters than temperature and you also have to consider the front end(s) which might fiddle with a prompt before it is delivered.
There are at least two ways of dealing with ransomware for DR purposes. You will have to accept that you will be losing some data and how much is up to you.
Option 1 is air gapped backups and this is the gold standard. The classic method is tape and you must remove the tapes from the robotics/drive and store them away from the production system.
Option 2 is something like Veeam's "Hardened Linux repo" and using immutability - This is a silver gilt standard method and you have to be very careful with this. A hardened repo uses a one shot admin account to setup a service for the repo - the username and password are not kept on the Veeam backup box after setup. Veeam will ask the service to set the immutable flag on backup files and only remove them after a set time. However, if the backup server has another way to get at the box housing the repos, then the flags can be removed.
Prevention is the best bet but is not infallible (its hard to prevent all future threats)! Air-gapped backups are the gold standard and a suitable LTO drive or two or a full robotics unit and a slack handful of tapes don't cost the earth.
When I somehow managed to decide to give up fags, 2 Apr 2018, after 30 odd years, my plan was to hit a vape to take the edge off. That lasted a few days when I realised I would be swapping one habit for another and I wanted shot of the whole thing.
So I did and I'm glad I did not go the vape route and I can't say it was easy.
What worked for me, was two strategies. Firstly: time period doubling (attaining goals) - one day, two days, half a week, a week a fortnight, a month, two months etc; secondly: I had a couple of "mantras" that I would repeat mentally whenever the cravings hit. They were "I don't want to die" and "I don't want to smell" - two classic symptoms of smoking, one likely in the future and one that is guaranteed permanently. It worked for me but I had to "say" those mantras a lot to start with but it did work.
I did try the conventional methods many years ago. However, to my mind, if nicotine is the addictive ingredient in fags, why don't patches/gum etc work 100% ? I don't think that the "habit" thing holds much water either - I had no problem with that side of things (*). Quite simply, there are other ingredients that are addictive in fags and they are worse than nicotine and there is no other way to ingest them without a burning fag.
I would say that for me it took about two weeks to know I was definitely on track but about 18 months to be able to feel I could (nearly) totally let my guard down. I had some bloody weird dreams at times!
(*) When the pandemic hit, I stopped biting my nails - a habit of over 50 years.
Well you are anon - how the hell does anyone call you?
I've migrated the vast majority of my VMware customers to Proxmox and all is fine.
Do bear in mind that Proxmox is just a distro on top of Linux and Debian. Its KVM and Qemmu on the virty side and that runs way, way, way more workloads than VMware could possibly dream of.
If you are holding it wrong then that is your problem.
I was a VMware fanboi for a good 20 years. I even have a VCP 4 from when it has ESX - RedHat n stuff or ESXi (basically Linux but less RH branding and a few bits n bobs chucked out)
vCentre has been an abomination for ever. Windows service first with a fairly rapid phat client, then a Flash thing, then rewritten when Flash died out. Oh who can't recall the thrill of two web GUIs that were both rather wank for a good five plus years. Oh, the laughs we had with certificates and password expiration policies suddenly appearing after an update and I read release notes. I could go on.
In the meantime ESXi lurches from horror to horror. I wrote a wiki page on how to avoid multiple PSODs (its purple, baby) when passing a GPU through to a VM. Dell's mildly integrated Intel (or was it Broadcom) NICs around 2015 vanished every now and then. Recently ... oh who cares.
VMware was always about being a MVP. Fuck that.
Qemu/KVM runs so many VMs in the real world that VMware is a rounding error in comparison and a very expensive one at that.
Notice how all that happens to messages is that they will be delayed - not lost! This is a catastrophic fail for a node in a matrix but it seems to be failing safe very nicely.
I remember a time when a bluey used to take a week to cross the world. My grandparents remembered a time when comms were impossible in many cases. I could go on ...
"My great-grandfather's set square is certainly old, but it still nails a right-angle every time."
I have a flat bladed screwdriver that belonged to one of my G^2 fathers. He may have inherited it. It is capable of delivering a massive amount of torque and the properties of the steel at tip and on the shaft may well violate everything I learned about metallurgy!
I can put a stilson on the handle and tap with a rubber mallet and old school screws generally oblige me. A quick squirt of WD40 often helps too. You have to push down on the screw hard enough to remain engaged but not so hard that you lock the (rusted) thread even worse. The tapping needs to deliver enough torque but without snapping off the end of the blade. Obviously you have to consider that the screw might be rusted right through and will leave some behind. The tip is work hardened to just the right point and the shaft is mild enough to absorb shock.
Now that's a skill you don't see taught in school. If it all goes wrong, get the angle grinder and the other modern stuff out 8)
ZFS is marvelous.
I use it on pfSense (FreeBSD) and Proxmox (Linux). Anecdotally, it beats the crap out of UFS for pfSense. I have a cluster of Dell servers and many ACPU based boxes and many Netgate hardware jobbies to worry about. Since ZFS became the default, I have not lost a filesystem, on mostly single disc setups. Those should have an equal chance of data loss when power fails but ZFS seems to fail safer.
Lots of lovely Proxmox (ex VMware) boxes. The ones without Ceph have ZFS, including my home systems. Lovely.
What isn't lovely is swapping out a failed disc in a zpool. Many of these systems have RAID controllers that I have converted to non-RAID so that ZFS can do its stuff properly. In RAID mode the controllers will do their periodic patrol reads and cast out discs that have been deemed to have failed. An amber light comes on and you swap it out - job done.
Its not exactly the end of the world doing the ZFS equivalent but working out which device is which can be a right old laugh. I can't help but feel that ZFS could notice a swap out of a similar disc as a failed one on the same bus/slot and ask if it should be used when you next run zpool. You would be prompted by an email from zed.