TacticalCoder

Press the button to turn off the power strip when you're not watching it.

The other solution for those who have a flat white wall (or are willing to install a projection screen) and which I used for years and years: use a dumb projector. There are plenty of perfectly fine dumb projectors and they're very cheap too (compared to the equivalent TV). This also made for the largest diagonal I ever had, in a huge living room, which was really great: basically turning the living room into a mini home theater. I painted the walls etc. accordingly, including a special paint for the white projection wall and black for the side and back walls. This was a better experience than any $$$ TV.

Sadly atm I can't use my projector for my wall ain't flat and I can't install a projection screen and I've got nowhere to put my projector: I could use a short-throw one but the place ain't mine and I'm not allowed to drill holes etc.

So atm it's... Pressing the button on the power strip that powers the LG TV. I know, I know: doesn't help much when the TV is on.

Next thing to do is separate all the junk devices (TV, smartphones [the ultimate spying device btw, much more than your TV], smartwatches, tables, etc.) from your proper stuff (e.g. your important PC running Linux/FreeBSD/etc.). I use different physical LANs (and a bridge/router) but VLANs would do too.

Many cheap home routers also by default offer at least a "guest" network: a perfect place the TV and your relatives' devices when they visit and want access. Set usage quota on the guest network too.

At some point all these devices shall use other networks than yours, so you'll be sorry out of luck anyway. But as long as they don't come with their own battery, pressing the button on the power-strip is some kind of giant low-cost "hack this" middle finger.

I'm all for banning vibe-coded projects because they are pure turds but I'm not for banning a codebase where AI is used to make the codebase better.

So what's the definition of a vibe-coded project?

I verify the code the latest SOTA/frontier models I pay for do generate and although I daily facepalm myself several times per day, it's still a tool allowing me to be more productive. That's why I've got countless examples of mediocre output to give: if people were to take a look, they'd notice the same.

The problem is the mindset "it runs, hence it's good" of fully vibe-coded crap.

I'm talking about stuff like "Add this information on the webpage next to that other info" and the model going on a tangent and pulling in new libraries to then do a computation on the server-side and then using client/server communication to send the info to then display it.

It works. It did what I prompted. We've got intelligent machines in 2026. Yay!

Except all that nonsense was totally unnecessary and the information was already on the client and all that was needed was to display it.

But the model saw a pattern: can you believe that? A pattern of information being exchanged between the client and the server. So of freaking course my prompt had to be solved by following the same pattern.

So what's that about? Full AI ban or ban of fully AI-generated turds? I'm all for the later but don't ban everything: AI can produce acceptable stuff at a rate that's making it a productivity enhancement tool.

Models have also only been getting better and better: as of 2026 fully vibe-coded projects are pure turds but what about 2028? 2030?

It's not because these things are introducing bugs in every five line of Bash code they write and generate what looks like an infinite amount of slop to solve things which should have required 1/10th of code that it's always going to be like that.

When people are talking about 50 EUR per month or 100 EUR per month of server expenses, are these problems that do require Kubernetes or is K8s overkill at these levels?

A very good friend of mine runs 1400 pods but his company (by that I mean the company he owns) spends something like 10 K EUR / month in servers. It's all Kubernetes and Talos (which another poster mentioned already) etc.

But at 50 to 100 EUR per month!?

Don't get me wrong: I'm excited to see people using Kubernetes with such low server budgets... I'm into VMs and containers, but haven't tried Kubernetes yet (despite my friend trying hard to convince me it's really not that complicated).

is the same exact reason the Kimi crowd is wrong now. And it's very obvious that they're wrong, but they have intense emotional blinders on.

On the very link on the top comment of this thread, which I repost here:

https://artificialanalysis.ai/models/gemini-3-6-flash

Kimi K3 is ahead of Fable 5 on several benchmarks.

So basically the angle went from "China cannot ever compete" to "China is six months behind" to "China is six weeks behind" to "China is six days behind but that's because they're distilling" and now you're saying "Yup sure, Kimi K3 is ahead on several benchmarks but you cannot host it yourself so this thing will go absolutely nowhere".

I mean: is it not a bit early to draw conclusions? It's been days since a chinese model is ahead of the very best / frontier US model on several benchmarks and you compare it to models who were clearly behind on everything.

Give it some time.

Anyway, I could just be emulating it instead (lol).

Yeah I'm wondering about that too. I'd say there's something special about using the real thing: have the old hardware working.

Ideally in a 100% real time-capsule mode: say a Commodore Amiga back then didn't have a "Vampire 8000 GTX Turbo 16-valves" accelerator. My machines are all pristine time capsules, not polluted by modernity.

But as you wrote: at some point something fails. And what do you do then?

I understand the appeal of "fake IDE disks" and "fake floppy drives" etc. but if it's to use stuff emulating part of the real thing, why not just use an emulator emulating the entire real thing?

It's not a criticism: for example I've got a real vintage arcade cab and more often than not the "PCB" that's plugged in is not a vintage PCB (even though I've got several of those) but a Raspberry Pi with a Pi2JAMMA adapter. And once a game is launched, you can't tell the difference: it's basically cycle-perfect emulation.

Something something about the ship of theseus and all that.

I'm very curious how much people will look back on this frenzy of PQC migration panic by 2050 when, my bet, there still won't be any remotely viable QCs.

The funny thing is there's, for the foreseeable future (decades), no viable QCs at all expect breaking non-QC cryptography.

And should breaking non-QC cryptography become even remotely possible, the effect is going to be that everything is going to shift, like SSH, to an hybrid solution with two algos: good old non-QC resistant algo and new fancy QC-resistant algo.

So basically QC will have achieved jack shit besides moving everybody to use two algos instead of one.

In 2060 you'll have one device able to use quantum computing to break some messages from 2030, incriminating old senile or long dead politicians, that nobody will give any shit about.

That's the weirdest thing about this whole thing: the only viable QC we hear about is "breaking cryptography" but as soon as that becomes a possibility, we'll all be moving (just like TFA and OpenSSH are showing btw) to QC-resistant algos.

Making QC non-viable.

So... Besides cracking old obsolete message, what the heck can QC be used for?

I mean: what good are those machines going to bring the world, besides cracking old messages nobody cares about and besides making every switch to QC-resistant schemes (like hybrid old+new algos)?

That's assuming we won't secure anything and we'll keep according approximately zero thought to computer security.

But from the look of it, at very long last, a great many people are beginning to now take security seriously. Suddenly they realize it's not just a teenager in mom's basement pretending to attack from North Korea but a near infinite number of AI that are the attackers.

I mean, yeah, we built worlds on PHP and JavaScript codebases and these probably don't stand a chance.

But it doesn't have to be like this.

I see AI as a chance to, at long last, have proper network security.

AFAICT cryptography hasn't been broken yet. There are still physical taps (physicall one-way only, undetectable) and honeypots out there. There are still some network where a single unaccounted for network packet is cause for inquiry (either a bug or an attack).

And for those who are not using proper security measures, they can now get the help of AI to set up better networks, to harden their bases.

I think people really downplay the impact of memory issues.

I don't think so. I've got my important data backed up on offline HDDs and SSDs and my most important data lives on a RAID system (ZFS) on a Xeon with ECC RAM but...

... there are simply a huge lot of checks everywhere making it so that bit flips on non ECC systems simply aren't that serious.

We live in a world of checksums and retries at every layer and distributed systems that are just about everywhere.

I went on for decades without ECC memory and nothing bad happened. Yes, I had the occasional bitflips. I did found out faulty mem using memtest (or weirder stuff, like some RAM stick identical to the other three, same chip everywhere, same sticker: but reported by memstick as being some other mem).

But I don't think that the "one bitflip cascading into corruption of your life's data" is something necessary to freak about.

It's not happening. Never happened to anyone.

If you (or other) are really concerned, you can use my scheme:

    - dsc09874.jpg

    - dsc09874-b3-73b0835221.jpg

About 1/3rd of all my files are "tagged" this way, directly in the filename: this survive different medium and different OSes.

I then have my custom rsync backuping script doing a dry-run and warning me in a "stop the world!" fashion if it detects a bogus checksum (either on the source or on the dest: during the dry-run any modification to a file with a checksum is verified and any checksum mismatch means "stop the world").

I then also have, in addition to my custom backuping script, another script that allows me to verify either all files with that naming convention or a percentage I pick of those files (so basically some random sampling).

It's not mutually exclusive with ZFS / a system with ECC RAM.

but not all our files are replaceable, and you forever and always might have corrupted files you are unaware of that got written during this time

I hear you but... How much do you really need, years later, a file that you haven't needed for years? That one file is unlikely to be a big loss.

Now I'm not saying you're wrong: I ended up with my naming scheme (appending a partial cryptographic hash to the filename) after finding a series of about 20 pictures that were corrupted. And on all the backups I first checked, they were also corrupted. I ended up finding a very old data CD on which that series of .jpg files were fine. And I decided "never again, I want such SNAFU to be detected in the future".

But truth is: with 40 000 pictures, life without those 20 pictures would still be fine.

https://arena.ai/leaderboard/code

(that's just one, there are others... and several are claiming that those saying "yes sure it's just as good, but it uses more tokens" ain't really true either).

And I don't think anyone is excited about China releasing anything: I think people like the idea of open-weights.

Their bread and butter and claim-to-fame will continue to be low-cost, near-frontier models, ...

K3 is beating Fable on several benchmarks. So I take it you don't consider Fable 5 to be frontier.

What are the frontier models?

They can invent it. They can build it. And it is only a matter of them before they can scale that last barrier of American hegemony- market it.

And at some point we'll see very capable chips coming out of China: Huawei, Baidu and Alibaba already have some stuff. I think it's only a matter of time before they come up with some AI accelerator doing 80% of the job at 20% of the price.

Yup and I read TFA and it's not clear: "The death and rebirth of my home server" and, somehow, his Raspberry Pi is his server!? Running RAID, I'm Mich (immich), Jellyfin, etc. all from a Pi is amazing.

Don't get me wrong: I've got like seven Pis, 4 NUCs (but only one "recent" N100 one) and even though two of the Pis are up 24/7 (one runs the unbound DNS and the other, well, serves as a router/firewall to isolate my main PC from the rest of the machines) my server, as in the machine that has disks in RAID (ZFS), runs VMs, containers (including Immich and Jellyfin), etc. is a Xeon with ECC RAM.

I had no idea a Pi 4 was that beefy.

Now of course booting off a microSD card: yuk (I know, I know, there are read-only trickeries that can be achieved).

The important stuff mostly lived in external HDDs and the configuration lived almost entirely in my NixOS configuration.

Yeah NixOS or what-have-you: if you have server, I'd recommend you have a plan B for when it fails. My little infra at home is mostly automated: it's all scripts provisioning the VMs, setting up cloudinit for the first boot, etc.

As long as I've got backups of my config/scripts and of my data, setting up a new server really doesn't take long (heck I did it not long ago, mostly to be sure the "death and rebirth" procedure of my own server was sound).

I personally prefer my 10+ years old Xeon with ECC as a server than my N100 NUC but then the Xeon consumes more energy... YMMV.

Then something has to be said about digital/discrete hardware, which is not dissimilar to programming, and then the alien (to software devs) world of analog chips.

Thank you for posting this, it reaffirms what I've been thinking for so long. There are many opportunities to retrofit old systems of all types with modern low-cost embedded technologies.

At my shooting range, where security is very important, we're replacing the old antiquated "fire stop" (when everybody on "shared" lanes has to stop shooting for there's at least one person going to change the targets) system with a modern system of electro-magnets and embedded devices. Mechanical safeguards where, by default, the 10 lanes (there two times 10 lanes, the 50 meters ones and the 100 meters ones) have to stop shooting and then an embedded system with a little controller for each lane that needs to be working to give the greenlight / "fire frei" ("free to shoot").

We mostly decided on the switch after a failure of the old system: the "free to shoot" system has recently been given even though there was still one person outside, changing the targets (nothing bad happened but this is a serious incident).

We asked for quotes for a system that is always "door locked" / red light / "fire stop", unless everything works fine and every lane gives the greenlight (each shooter has now one badge).

Same thing as in TFA: cost in hardware is minimal compared to the quoted prices. Like basically 1/50th of the quoted price.

I see a number 4770 and 4790 still doing the job they were bought for. I expect the same for 12th gen when they're 13yo. Not so for anything between 4th and 12th.

What do you mean? My Core i7-6700K is still running strong on its Asus Z-170A mobo (TFA mention a Z370-A, mine's a bit older but was one of the first mobo to offer a M.2 NVMe slot). It even runs very modern games just fine on whatever not-so-powerful GPU that's in it: it's the PC my daughter uses to play (I did put PoP OS! and Steam on it but I'll swap to another distro if needed).

I switched to an AMD 3700X and then to an AMD7700X but I don't see anything wrong with my 6th gen Core i7.

Obviously very depended on how you value your time ...

More like how you approach life.

Assembling furnitures is a very fun and zen thing to do IMO. I know a least one surgeon who would gladly bill more $$$ but there are simply hours when the hospital is closed: assembling a furniture with your kid after a day's work is a cool and relaxing thing to do.

Friends of mine, about a year ago, asked me to help with some technical stuff on the first day they moved their startup in their new office. When I arrived, the desks hadn't even been assembled yet (they were late on schedule).

I helped them assemble the desks: I find that highly relaxing, gratifying and relaxing to do.

Many moons ago I did that too in southern california: physically building the desks, unpacking the PCs and monitors, hooking up everything (back when there were still, btw, people plugging the VGA cable instead of the HDMI one for monitors that came with both), running the ethernet cables, configuring the PCs, configuring the Linux server.

Doing it all. Then configuring the Linux servers.

It's like painting a place you move in's walls: it's got nothing to do with "how you value time".

I lived in four countries. We'll certainly move to a fifth one one day. I'm excited at the idea of going to the hardware store, buying paint and screws and nuts and bolts and whatnots and preparing the place.

And assembling furnitures.

That might be an incomplete date point, so using the next most recent value 1226, StackOverflow has lost 99.41% of its activity.

The biggest problem was highlighted from day one: StackOverflow was bound to become "DeadOverflow". It couldn't possibly work because the entire notion of having one correct answer to a question in a domain that constantly evolves was broken.

What killed the site was "Ever relevant question" showing up in a Google search (pick other search engine) and pointing to a "Correct and accepted answer" that was wrong with then another answer with more upvotes but which was... Already outdated.

That's why people coined the term "DeadOverflow": the very way the site was conceived would inevitably lead to dead answers.

And people pointed that out early on. Nobody listened, many of the OGs who created that company made a nice exit.

But SO was destined to eventually fail.

The entire karma/gamification/clique of users gate-keeping was bad but the issue of dead answers / DeadOverflow simply couldn't be solved, so it doesn't even matter.

Losing 99.41% of its activity is brutal but it's honestly surprising that a concept that couldn't possibly work even survived that long.

1. Your OS installs malware ...

Your OS is malware.

We're talking about Windows here.

... stream it over the Internet to my phone, that's connected to Android Auto. That's how to play a CD on a modern car!

Sadly when you stream music over Bluetooth there is lossy compression taking place.

People can bitch all they want about "320 kbps mp3 using a modern encoder being indistinguishable from the lossless source", the fact is: mp3 were nice 30 years ago in the Napster day (when I had my ADSL Internet connection).

It's 2026 and I play lossless music, including FLAC files I bit-perfectly ripped from my own CD collection (and the rips have been verified online with a DB of rips, so I now they're bit-perfect) and including lossless music from proper music streaming platform (I pay a Qobuz subscription: amazing quality, huge offering, but shittiest music discovery platform ever btw).

Really: it's 2026 and if people like to listen to lossy music, more power to them. I don't, and hence I don't stream audio over Bluetooth.

My car both takes CDs, memory sticks (mp3 and WAV, no FLAC) and has got its own memory. I've got a few hundred songs stored in the car: they go losslessly to the car's DAC.

High-end luxury car and it's the type of car where people who've never been in one say: "I never knew music could sound that good in a car" (which btw is funny to witness when some car youtubers try a car with an amazing soundsystem).

Now of course it's still a car. Don't get me started on home stereo because I've got that too (and when you know what you're doing you can get amazing audio for a really very reasonable price: for example I won't put 30 K in a McIntosh amp where a Yamaha amp costing 1/30th of the price will do).

Call me an audiofool and... Enjoy your lossy Bluetooth streaming adding a layer of lossyness on your already lossy mp3 on your car stereo system which sounds like crap anyway.

I can easily read both but here's the funny thing: with my reading glasses on, I first see "Sorry robot". If I remove my glasses, I first see "Happy human".

Which makes me think this is one blurr filter away from being trivially read by any model.

Very cool.

... but it's a bad fiction to perpetuate that any of this is anything more than tarot card reading

Most people's issue with AI-generated llmish however is not that it's AI-generated. It's its insufferable tone.

So if we get to a point where we have to read tea leaves (an image you seem to appreciate) to determine if it's llmish or not, we'll have won by then.

Really: it's that full-on asshole tone I (and many others) want to see disappear from blogs, comments, LinkedIn, etc.

Yup some here are in denial but what many said would happen did just happen. They're not "six months behind": the model is totally SOTA. Cheaper, faster and they don't just crush Sonnet 5 and Opus 4.8: on 6 of the 14 benchmarks they posted Kimi K3 is in front of Fable.

Of course the shills are shifting their tone: this thread as devolved into "sure yup it's totally SOTA but it sucks because it'll use more tokens than Fable to do the same task".

I take it that's the new tune we'll hear for a while. Oh well, at least we won't have to suffer the "they're six months behind, so they're totally useless" anymore.

P.S: I'll make a prediction... We'll hear the "buuuuuuuuut it uses more tokens for the same task" for a few weeks, then we'll get Fable 5.1 and those same posters are going to post "Fable 5.1 is so much ahead you're missing out if you're still on that piece of turd that Fable 5 or K3 is".

Where are the voices of reason?

My wife got an email from a new hire (now even a new hire yet: she's still on a trial basis), a 23 years old, where she explains that she doesn't want to use AI. That she doesn't like what AI does. On a funny sidenote: the email is obviously 99% llmish, which is hilarious.

That's one extremity: crazy people who refuse to learn a new tool.

Then on the other extremity you have the even much crazier ones: those who believe they've got an intelligent machine that is going to solve all their work problems during the day and then, at night, that is going to enlighten them by revealing them who god really is.

Where the heck are the reasonable people who use AI for what it is: a tool that can be extremely helpful at times and extremely sucky at other times but that is still, on average, a time saver?

At this point there's more patch than rubber visible on the the old Windows tire.

I don’t agree that “everything is a file”. Files are arrays of bytes. For an LLM, everything is a vector of tokens/embeddings.

And yet, as of now, LLMs have a hammer (Bash / command line utilities) and every problem they have looks like a nail.

If there are people around you who are non-techies that are using Claude Code or similar, you'll hear them ask "what the heck is cron?" and "why is it talking to me about Bash again?".

At some point we may have LLMs working on their own output using a set of tools that'd be the equivalent of Bash (or any other terminal prompt) + command line utilities manipulating not files but tokens/embedded vectors but as of now, it sure looks like everything is a file, especially to LLMs.

Yes and a 10 year old Xeon is going to be a v4 (not a v2 as in TFA) and it's going to have DDR4 ECC, not DDR3 ECC.

I've got a 14 cores / 28 threads Xeon from 2015 that I use as a server at home (ZFS / VMs).

It's really a sweet machine.

For ricing I've got a semi-recent AMD 7700X / DDR5 RAM (from 2023 ?) which is my main machine but the real deal is my old and trusty 10 years old Xeon server.

DDR4 ECC is pricey too atm but a 10 years old Xeon is basically free now.

A 20 cores / 40 threads costs maybe 20 USD (for just the CPU). Slap that in a $100 old HP Z440 workstation and you're good to go for quite a few workloads.

Mine is only on when I'm at my computer: it's not turned on 24/7 but more like 8/7 so the entire "but it consumes energy" point is moot.

Totally overrated.

The Acquiescence Effect (also known as Acquiescence bias, Yea-saying or Agreement effect) is an observed tendency for people to choose positive answers (such as Yes or Agree) more frequently than negative answers, regardless of the question, in the range of 10-50%.

I trained my kid from a very young age to learn that "no", not yes, was the most powerful word in the dictionary. Never obey. Say no. Refuse.

I never feel obliged to do anything. You have zero power on me. Zero. I'll do a performance act when people with actual authority talk to me (like, say, policemen) while keeping at all times in mind where their power ends (which is that basically after they're done talking to me, it's over and I won't ever talk to them again).

People should learn to use the most powerful word in the dictionary, which is "no".

As Rage Against The Machine nicely put it: "Fuck you I won't do what you tell me".

I don't know about that but finding excuses for the scum of this earth is certainly not a solution.

Take Europe for example: nobody dies of hunger in Europe. And yet there are plenty of thieves. People stealing tens of thousands, hundreds of thousands, millions of EUR aren't doing it to "feed their families".

Think of the situation today. Think of the victims today. Instead of thinking of tomorrow's hypothetical situation where supposedly all the honest fathers out of work would join the crime syndicate, think of today's victims.

Projecting your own insecurity about the future to excuse scummy behavior by the scum of this earth is of no help.

There are people, right now, who have a roof. Who have a family. And who are fucking scums stealing the hard earned money of others because they choosed the easy life of crime.

Zero tolerance for such motherfuckers. I care about the victims and you should too.

I'll fix this entire HN thread for you all and you'll thank me later:

    s/bug/backdoor/g

I use Telegram only for groups with people I know. I've got zero issues with boths.

Are you using public channels? (are those even a thing with Telegram?)