HNHacker News
TopNewBestAskShowJobs

arghwhat

9,034 karma · joined January 12, 2017

no.
submissionscomments
arghwhat··on GCP Outage
Redundancy ≠ immune to failure.
arghwhat··on The Danish Ministry of Digitalization Is Switching to Linux and LibreOffice
> Usually the biggest hurdles are Active Directory, deep Excel trenches and reliance on Outlook.

It's mostly a mental block for the corporate IT department. People are already familiar with other email and calendar apps (e.g., gmail, apple mail, ...), and Active Directory only matters if you're running a lot of internal infrastructure feeding off it, usually through standard protocols like LDAP or OIDC.

> How are Danes submitting taxes?

Regular people don't submit taxes, they only make changes to the report if necessary once it comes out (e.g., to fix a missing deductible, such as commute distance deductible). Most stuff, including deductibles, are automatically reported.

All tax, including personal corrections and corporate tax, is done through an online web application. We do depend on smartphone apps for Android and iOS, but there's no Windows desktop apps or applets anywhere - at least to the public.

arghwhat··on The Danish Ministry of Digitalization Is Switching to Linux and LibreOffice
While I won't comment on whether the move made is smart:

> 1. How much is compatibility with outside users and past documents necessary?

Not an issue, document compatibility is a solved problem and any issues experienced would be minor (e.g., a missing font or broken animation leading to slightly different presentation).

> 2. What is the system management story?

Corporate "system management" solutions are entirely broken on Windows, especially upgrade and software bundle management. It actively causes more harm and security issues than it solves, and needs to die.

It's dumpster fire of technologies that get stitched on top of each other, all conflicting and ultimately training all employees to ignore all signs of malicious intervention: How do you expect a user to sceptical of questionable popups/browser hijacking/similar, when your setup involves training users to ignore cmd.exe windows popping up randomly and to always interact with highly inconsistent, constantly changing and overall questionable popups appearing at any time of day asking you to do weird things (installing apps you didn't ask for, updating, rebooting, upgrading, often with countdown timers for making decisions)?

Should you have the ill-advised desire to bring the worst and most defective parts of Windows IT management to Linux, there's a handful of big vendors providing similar solutions there. So "not an issue", other than it being a terrible idea.

arghwhat··on He has powered his house for 8 years using laptop batteries
I don't see how a short consultation on the matter would exceed 3-4 digits. It's maybe an hour from a specialist to see what is needed. Saving your bacon later on is worth some cost.

Worst case, you might also have existing recommendations or fire code/regulations for battery energy storage systems in your area, and while those specifically may be overkill they give an idea of what is considered necessary. Maybe you could consult your local fire department, depending on how things work in your area.

arghwhat··on He has powered his house for 8 years using laptop batteries
When you only have one battery cell it's also hard to screw up battery management, and phone batteries are rarely driven past their prime (no one wants a phone with 50% remaining capacity). The odds are not zero though, as the battery can develop internal shorts from imperfections like dendrite growth over time.

The odds get much worse when you have tens or hundreds of kWh comprised of hundreds or thousands of cells, especially in OPs jank setup with improper battery management and cells that are all already at the end of their life and likely have significant internal damage already.

arghwhat··on He has powered his house for 8 years using laptop batteries
Your home residential battery is not an entire parking garage full of EVs, so looking at their pipes and pressures is not relevant.

Those pipes are dimensioned to be able to suppress fire in every parking booth of an entire parking garage full of EVs, each of which packing more of a punch than most residential battery installations. While it cannot handle every car being on fire, it needs to handle any car being on fire when it goes off.

This is orders of magnitude more powerful than what you need for a small residential setup. Have a professional evaluate what system is best - if they suggest a sprinkler system, which they might, I am sure they would understand the capacity of your residential water supply and suggest upgrades (new piping to mains or buffer tanks) as needed.

arghwhat··on He has powered his house for 8 years using laptop batteries
I think you're confusing suppresing the fire and extinguishing it. You are not trying to stop battery cells that are currently on fire from burning - you will just have to wait for those to be out of fuel. The purpose of fire suppression is to stop new things from heating up by the fire past the point of combustion, catching fire and thereby spreading it - be it more battery cells, construction materials, you name it.

Modern EV fire suppression systems for parking garages use high pressure water mists to contain the fire to a single vechicle for example. This keeps neighboring cars cool and avoids them catching fire, and cooling down the burning vehicle may avoid spread to more battery banks or other flammable materials.

arghwhat··on He has powered his house for 8 years using laptop batteries
Battery fire suppression systems often use water, and firefighters will also certainly apply water. Just don't walk up to the cell with a bucket of water.

Water can and is used to cool batteries during battery fires, and more importantly, to cool the surroundings so that the fire cannot spread to, say, neighboring banks/cells or construction materials. Modern EV fire suppression systems for parking garages use high-pressure water mists to avoid the fire spreading to neighboring vehicles for example.

The recommendation about oil and water focuses on a larger container of liquid fuel that is on fire at the surface and heated far past the boiling point of water, such that dumping a large volume of water onto it all at once causes it to immediately boil and explode, spreading large amounts of oil as a mist in the air, which both spreads the fire and causes a much more violent combustion. A water-based fire suppression system (not a guy with a bucket of water), or a firefighter with a water hose, can absolutely extinguish such a fire.

Hacking together a sand container of acrylic may well do nothing to limit the fire while simultanously giving it more fuel (acrylic) and pathways to spread to (whatever the acrylic is near).

The point is: Don't hack together a fire suppression system, leave that to an expert please.

arghwhat··on He has powered his house for 8 years using laptop batteries
The banks are "air gapped" by just a few cm, are all interconnected with excellent heat conductors (the bus bars), is insulated with flammable plastic, and sit on a flammable shelf in a flammable shed.

There is no fire safety there. Let's just hope the shed is very far from a house, fences, trees, bushes, or anything else that could catch fire and close any "air gaps" and cause harm.

arghwhat··on He has powered his house for 8 years using laptop batteries
Acrylic is itself flammable, massive amounts of sand would be required to fully submerge and cool all cells (and therefore hefty acrylic sheets less likely to give), and it's likely that the sand will not spill correctly and evenly.

Please only use properly designed and tested fire suppression systems, as hack jobs might not help at all and do harm from the false sense of security leading to lack of actually effective mitigations.

arghwhat··on He has powered his house for 8 years using laptop batteries
If the batteries catch fire the acrylic bottom will melt and automatically dump the sand onto the fire.

This is a questionable setup though. You'd need massive amounts of sand dumped evenly, which requires more design and verification. Acrylic is also itself flammable.

A basic water based fire suppressor would not extinguish a battery fire but it will cool it and the room, limiting spread.

Let the experts design this kind of thing.

arghwhat··on How Does Claude 4 Think? – Sholto Douglas and Trenton Bricken
> If an AI correctly guesses the outcome of a dice roll, it does not mean it knows the outcome in advance. Being correct is a necessary element of knowledge, but far from sufficient.

If we "know in advance", why wouldn't the AI?

If the AI doesn't "know in advance", why would we?

When I'm asked a question, the response and thoughts around it emerge only at that point, and from thin air. I certainly don't trawl an internal, constant knowledge bank of retrievable facts, and the time of day or preexisting conversation may affect whether or not I successfully "know" something in a particular conversation, despite having learnt it. At the same time, some things I "know" will just be wrong - the memory distorted or mixed with another, the whole thing imagined, or even just a failed recollection. And no, I certainly don't know if it has gone wrong unless it happens to result in e.g., contradictions I notice myself.

These kinds of justifications for why AI doesn't "know" always imply that the process of human thought emergence is well established and understood and easily differentiated, which just isn't the case. There most certainly are more differences than similarities between AI and the human mind, but I dont think the concept of knowledge is necessarily one of them.

arghwhat··on How Does Claude 4 Think? – Sholto Douglas and Trenton Bricken
That does appear to be a better term for both humans and LLMs. Not sure who coined the term "hallucination" in the context of LLMs, which that suggests problems with perception rather than memory...
arghwhat··on How Does Claude 4 Think? – Sholto Douglas and Trenton Bricken
I'd love to hear your definition of human thinking.

No one else knows how that works, so it would most certainly be worth sharing, especially what the discriminator for whether something "thinks" would be.

While avoiding being "laughed out of the room", of course.

arghwhat··on How Does Claude 4 Think? – Sholto Douglas and Trenton Bricken
> because they don’t have knowledge like humans do, they rely on pattern-matching, not understanding.

I don't think this comparison makes sense - humans do not have knowledge as a unique thing differentiating us from LLMs. "Knowledge" is just the ability to produce relevant/"correct" results in some interaction, and some of us have acquired experiences that involve dealing with a particular subject and other knowledgable people within that subject for years or even decades, granting us (very fallible!) "knowledge" within that specific area.

Humans hallucinate answers, even within their area of knowledge. Our memory is imperfect, so we're creating an answer based on bits and pieces, while anything missing is subconsciously glossed over and filled in with something plausible. It's just that the more experiences you have in an area, the better your hallucination. Those weird moments when your parents assure you of some fact that challenge the limits of just how wrong something can be would then be an example of hallucination in the extreme polar opposite end.

(Note that I am not implying that the human brain works like an LLM, but rather just challenge the concept of "knowledge" being fundamenetally different from LLM behavior.)

arghwhat··on Denmark to raise retirement age to 70
1. Denmark has a special retirement system for people who started work life earlier than normal, allowing them to indeed retire after some 45-48 years of work. All these rules are broken though.

2. These ages only affect a minimal government pension, and a massive tax saving on cashing out a specific, government approved retirement investment. If you wait to start receiving payments on that till retirement age, you avoid paying income tax of it. You can start earlier with worse taxation, or use regular investment types altogether. The overcomplicated tax system and its incentives are quite broken though.

3. Danish politicians still have a retirement age of 60 and a massive retirement package, so they can't deem retirement age and available funds to be that important...

arghwhat··on Does Earth have two high-tide bulges on opposite sides? (2014)
> Edit: fundamentally, if hacker news has taught me anything, it's that "downvote = makes me feel bad and doesn't want to answer questions". The entire concept of democratic news aggregation was a lie.

Guidelines:

> Please don't comment about the voting on comments. It never does any good, and it makes boring reading.

That it feels bad to not win the popular vote does not make democracy a lie, and there's no surprise in not winning favor when blanket discarding the current topic and describing it as "hubris", while not adding any new or constructive information.

arghwhat··on Tallest Wooden Wind Turbine
https://en.wikipedia.org/wiki/2025_Iberian_Peninsula_blackou...

The root cause is not known, but Spain was producing excess power (primarily solar) at the time around the disconnect. Some fluctations were seen, then supply started to disconnect from the grid in Spain, leading to sudden loss of 2.2GW of power. In Spain, automatic load shedding then happened to try to recover, but it was too little too late as neighboring countries detached from Spain to protect their own grids.

Nothing about this sounds like an issue with renewables.

arghwhat··on Tallest Wooden Wind Turbine
"Uneven electricity output"?

The variation on output is over a matter of hours (wind powerful enough to spin entire wind farms is not something that comes one second and is gone the next), and large grids with import and export capabilities are largely self-regulating.

Cost fluctuations in the electricity market regulate whether e.g., power storage sites will charge or dump power, whether district heating plants will source more heat from giant electric kettles, when EVs will start to charge, when private smart water heaters will preheat, when people decide to schedule washing machines and dishwasher, whether offline fossil fuel power plants will be fired up to sell as the rate becomes more lucrative or shut down as power becomes too cheap, whether any "idle" plants will throttle up or down, and whether windmills will engage brakes and turn away from the wind or release brakes and turn into it.

Power grids have also always had the ability to load shed by dropping customers off the grid, starting with factories that have special agreements, in case the combined local production and import is insufficient, and can detatch from neighboring grids and countries if there are import/export issues that could destabilize the grid.

The grid needs to change when supply or load conditions change significantly (e.g., every house in a city suddenly having an EV or heat pump, every house in a city suddenly having solar cells and supplying a ton of power, a power plant or wind farm being built somewhere power has not previously been routed), and can be optimized (e.g., power storage, smart load scheduling), but that is entirely orthogonal to windmills.

arghwhat··on 1,145 pull requests per day
In finance, not making mistakes is not at all critical, and mistakes happpens all the time. Regulatory compliance is critical, as it provides a thorough legal defense even in case of failure.

The lack of efficiency in finance (or pharma for that matter) is not driven by a wish for quality, but purely from a fear of stepping outside regulatory compliance with no individual wanting to be responsible for any process optimization that could later be seen as non-compliant.

Younger companies on the other hand might realize that compliance controls are, in fact, something the company defines themselves and can structure and optimize however they'd like, allowing for both high throughput and compliance. It's just hard to implement in older companies where half the staff would fight back as their role exists purely due to poor processes and their overhead.

arghwhat··on Claude 4
> Open-access publishing implies the authors are paying the cost of publication and its popularity in STEM reflects an availability of money

No it implied the journal not double-dipping by extorting both the author and the reader, while not actually performing any valuable task whatsoever for that money.

arghwhat··on The scientific “unit” we call the decibel
The problem isn't really the ratio, but the use in arbitrary contexts that require a lot of pre-existing knowledge. The reference value is sometimes a rather arbitrary value in an arbitrary unit, neither of which is communicated by the "dB" unit suffix.

The SI way to write `10 dBm` is to write `10 dB (1mW)`, clearly communicating both the power level and the reference point and unit. This ensures that you do not have to just memorize a bunch of decibell suffixes and their magical reference values.

arghwhat··on Edit is now open source
No. Adding superfluous features to an intentionally minimal baseline program defacto reduces it's quality by making it less suitable for its task and more likely to have defects.

There's decades worth of options for fatter editors out there. Tools like nano (and now edit) is ubiquitous because they are not such editors, and people need a reliable baseline without such features.

arghwhat··on Edit is now open source
I prefer SSH because it fits into a standard workflow with far larger ecosystem (my file manager and editors can browse SSH directly, traffic can easily be routed over SSH in either direction), and is already what I'd use for the majority of servers and machines.

psremoting would be limited specifically to remote PowerShell on windows, which is comparatively limited. Even if I was purely a Windows admin I'd consider the SSH approach more powerful.

arghwhat··on Edit is now open source
SSH is ubiquitous, lightweight, and now works for all machines equally.

SSH is integrated with everything and can be used to manage, transfer files/mount remote folders, forward ports in either direction, or proxy any network traffic through the remote end for debugging/access. You can open a remote folder directly in vscode, Zed or similar over SSH.

RDP to a standard Windows server (this was previously called "core" - having a desktop environment is a non-default add-on now) is silly as you're RDP'ing to solely see a cmd.exe window that asks you to use console-based configuration (sconfig.exe).

RDP to a windows server with "Desktop Experience" is silly because it is comparatively sluggish, streaming a (likely unaccelerated) video feed of heavier GUI applications vs. sending just a few text strings back and forth during editing.

Not to mention that containerized windows has no graphical stack and requires a text-centric workflow.

arghwhat··on Edit is now open source
... Yikes.
arghwhat··on Edit is now open source
This is a console application, it does not manage the keyboard directly and there is nothing windows specific here.

All existing editors do these things well already, nothing that required writing yet another editor.

arghwhat··on Edit is now open source
More features ≠ higher quality.
arghwhat··on Edit is now open source
That just feels like scope creep. Notepad would also be neat with some better keybinding support, plugins, lsp, ... - but then it wouldn't be notepad, and theres countless not notepads out there.

Tools like nano (well, pico) exist to provide a reliable and always available minimum feature set. If you expand it, then you end up with something that is neither the minimum nor capable enough to sensibly compete with fully fledged alternatives.

arghwhat··on Edit is now open source
I'm actually not sure, but not because I believe there are many of the latter.
← PreviousPage 11 of 34Next →