HNHacker News
TopNewBestAskShowJobs

grumbel

1,782 karma · joined April 16, 2015

submissionscomments
grumbel··on Lilygo T-Deck: 2.8-inch IPS LCD display, mini keyboard, and ESP32 processor
There are numerous $50-100 Rockchip-based GameBoy-like emulation handhelds on the market, such as the R36S. The R36S runs normal Linux (ArkOS/Ubuntu), has an audio jack along with numerous buttons. It's missing a touchscreen and WiFi, but has an USB port to plug in a keyboard or WiFi adapter (USB port is somewhat current limited, so it won't work with all).
grumbel··on There is an OOM kill count in Linux
More RAM. It's the only thing that can reliably prevent it. Magic SysRq key[1] can trigger the OOM manually, which can sometimes rescue the system if you notice it soon enough. But if you run into the situation regularly under normal use, not programs running amok, more RAM is a must.

If you can't upgrade the RAM, zram[2] is also worth a try.

[1] https://en.wikipedia.org/wiki/Magic_SysRq_key

[2] https://en.wikipedia.org/wiki/Zram

grumbel··on There is an OOM kill count in Linux
You have only about 10sec between the system getting slower and the system locking up completely. If you manage to hit the Magic SysRq key combination to trigger OOM manually, that can save the system, but you have to be quick.

[1] https://en.wikipedia.org/wiki/Magic_SysRq_key

grumbel··on The Web Is Fantastic
> never went away.

It might still be hiding out there, but when search engines never point you to it, social sites don't allow links to it and even your browser lacks any advanced way to deal with it (recommendations, notifications, editing, uploading, etc.), how are you ever going to find it? The amount of normal open web sites I tend to interact with is approaching zero.

Doesn't help that the retention on the open web is pretty bad. All links turn into 404 within a couple of months, so that little bit of interaction I still have with the open web happens mostly through archive.org.

It might not be dead yet, but it's certainly not in a very healthy state either.

grumbel··on Nintendo Network shutdown – The beginning of the end
Steam will drop Windows7/8 support January 1, 2024[1], Vista and XP support got dropped a long time ago. Meanwhile they still sell Win95/98 games on the Steam store, which won't run well or at all on a Windows10/11.

So not exactly perfect either as far as backward compatibility is concerned.

[1] https://help.steampowered.com/en/faqs/view/4784-4F2B-1321-80...

grumbel··on Spare a Thought for Julian Assange Spending His Fifth Christmas Behind Bars
Snowden spend the last ten years in exile in Russia, that's not that much of an improvement.
grumbel··on Ask HN: What's the coolest non standard application of LLMs you've seen?
Worth pointing out that most of that video is fake[1]. Though it and its debunking video are still a great example how to make entertaining fictional content with a little help of AI. It probably won't be too long before somebody builds something like that for real, similar AI mods[2] for Skyrim are already out.

[1] https://www.youtube.com/watch?v=bmqUUb80ApQ

[2] https://www.nexusmods.com/skyrimspecialedition/mods/98631

grumbel··on String pinsetters are upending bowling
Must be a US specific thing, since we had those pinsetters in nine-pin bowling/kegeln in Germany for many decades, e.g. a Kegelbahn from the 1960s[1]

[1] https://www.youtube.com/watch?v=geLZqO4ySoQ

grumbel··on Sony debuts first PS5 controller for disabled gamers
> Compared to the adaptive xbox controller this thing looks absolutely useless to me.

The XBox adaptive controller doesn't have sticks either. The two big touchpad-looking things are just two big digital buttons without any touchpad functionality. The Xbox adaptive controller really isn't much of a controller by itself, but just a large hub where you can plug accessories into via USB for sticks and 3.5mm jacks for buttons.

grumbel··on Sony debuts first PS5 controller for disabled gamers
Those effort are always a bit iffy since those very same companies go to great length to lock third party controllers out of their ecosystem, as well as fail to support even their very own controllers from previous console generations.

If real accessibility would be the goal, they could just open up their ecosystem and make proper use of USB and Bluetooth so everybody can build their own controller and keep using them across console generations as well as consoles from different manufacturers.

grumbel··on PeerTube v6
> Why not propel a platform you don't have to fight against?

That does not exist and fundamentally can't exist. A platform is by definition a thing I have no control over. Somebody else runs the thing and they can do with it as they please. And even if a platform is nice to use today, it sooner or later will get enshittified.

The only actual solution is to move as much data to the client and let them decide how to handle it. Especially when it comes to metadata, that shouldn't be that difficult. There is no reason why something like channel subscription can't be handled locally.

grumbel··on PeerTube v6
It's doesn't really matter what the service is doing when you aren't using the service in the first place. Right now I can use `yt-dlp`, download videos from Youtube and watch them in a video player of my choice. I never have to interact with the Youtube platform directly for this, it just acts as dumb server that hosts `.mp4` files. Same is true for a lot of other services.

The missing part is a good way to discover those videos in the first place. Subscriptions and recommendations are not something your average Web browser or search engine provides, yet are really important for discovering content. That's the kind of features I'd like to see implemented in a service independent manner. RSS went that direction, but not far enough, as it generally depend on the service itself to provide that information, instead generating it automatically from information found via webscraping.

grumbel··on PeerTube v6
Yep, efforts to replace Youtube (and the rest of BigTech) should be spend on building better search and cataloging of videos across all the Web instead of focusing all effort on one specific hosting technology that nobody uses. I'd be much more inclined to use a service or software that improves on Youtube by including other sources of videos, then one that tries to replace Youtube with their own set of (drastically inferior) videos.

Miro[1] tried that a long while ago, but I haven't really seen much else going that direction since then.

[1] https://en.wikipedia.org/wiki/Miro_(video_software)

grumbel··on LLMs can't self-correct in reasoning tasks, DeepMind study finds
Asking an LLM about letters is like asking a human about the individual wavelength that make up color. They might be able to infer it from text they read, but their sensory system is simply incapable of detecting any of that directly.
grumbel··on LLMs can't self-correct in reasoning tasks, DeepMind study finds
One big difference is that LLMs don't do loops, new words are produced in a fixed amount of steps. They can't go "give me a minute", think for a bit in the background and come back with an answer. That's why you can often get better answers when you force them to go through the individual steps, as that allows them to use the prompt history as working memory.

That said, this all seems like a 'fallacy of composition'. Humans are not LLMs, so much should be obvious, but at the same time concluding that they are completely different just feels wrong. The mistakes LLMs make feel very similar to what humans do when they don't have the time to deeply think about a problem and just give you the best guess that pops into their head. Humans will have other systems on top that allow them deeper reasoning, but the language generation really doesn't feel all that different from what LLMs do.

That aside, humans interact with the world, they get instant feedback on what of their predictions is right or wrong. LLMs are stuck with just static training data that might simply not be enough to develop higher level reasoning skills.

grumbel··on Amino – The Public IPFS DHT Is Getting a Facelift
> I personally wouldn't keep my private photos on a public network, but everyone is different.

Well, that's exactly the problem, isn't it? IPFS could be extremely useful for local and private storage, as it provides a network file system with proper directories, an optional HTTP interface, content addresses and an fuse implementation to mount it on Linux, along with automatic distribution and caching of the data. Those are all excellent features that I haven't really seen in any other system.

But the actual support for local or private hosting is basically non-existent. On IPFS everything is public all the time. The whole thing is way to much focused on being a globally spread protocol, while it neglects the benefits it could provide on the local PC, by just being a file format.

What I am missing is something like Git build in top of IPFS hashes. Something that allows me to manage my files on my local PC without any of the networking, but with the content addressing. Something that allows me to quickly publish them to a wider audience if I desire, but doesn't force me to. Or even just something I can use as a way to access my local files via content address instead of filename.

grumbel··on -Werror Is not your friend (2017)
Enable -Werror for the reproducible builds, leave it off for the rest. That should solve most of it.

The biggest and rather frustrating issue left is that warnings are heavily tool depended, not just in that new tools will produce new warnings, but in that different tools will produce incompatible warning. The workaround that makes the warning go away in one tool, will create a warning in the other. You can end up with a lengthy back&forth before you find something that makes all the tools happy and you frequently run into situation were you are really just pleasing the tools, not fixing any actual issues.

grumbel··on How to Use Monadic Operations for `std:optional` in C++23
The & captures the `userId` by reference and makes it available in the lambda body.
grumbel··on Is Math Real?
How do you agree on those root principles? Don't you need a set of even lower level shared rules and ability that allows you to agree on anything in the first place? If I show the paper with the rules to a dog, I don't think they can agree or disagree with me in any meaningful way, despite otherwise being reasonably intelligent creatures. Could there be aliens that are vastly different from us that can make rules that we would be unable to agree to? Could there be aliens similar to us, that would have different means to agree that are incompatible with ours?
grumbel··on Valve is a wonderful upstream contributor to Linux and the open-source community
> I never thought we'd see a no questions asked return policy for digital games and yet, with steam, it's a fairly simple process.

That was however not Valve's idea, but only came in response to EA Origin offering a return policy in 2013, Valve follow up in 2015 with their own.

One of the very rare cases where the competition was a little ahead for once.

grumbel··on Forty years of GNU and the free software movement
AGPL is only concerned with source code, it doesn't address how to handle a user's data in any way. Giving you the source, while keeping your data under locks is perfectly fine under AGPL.

Meanwhile the GDPR has the concept of "Data Portability" and requires every service to allow users to export their data, so that they can move it somewhere else. As far as I know, there is no Free Software license with similar requirements. They all stop at the source code, which just isn't enough in a modern SAAS based world.

grumbel··on Forty years of GNU and the free software movement
The issue for me is more that the FSF hasn't kept up with what is happening in technology. Software these days is often a secondary concern, as you aren't the one owning the computer, instead it is all running remotely and it's really about the data and control of said data. The licenses the FSF offers, including the AGPL, are inadequate to address this.

Over in Europe we got the GDPR, which does actually address a lot of data related issues quite well. But as far as Free Software licenses go, there still isn't any equivalent or even much discussion on how to attack the problem in the first place.

FSF feels stuck in solving 1980s problems, while having little to say about 2020s problems. "Don't do it", "self-host" or "run locally" doesn't cut it when a lot of software fundamentally has to be distributed in some fashion.

grumbel··on Philosophy of Coroutines
The one big issue I have with coroutines is that they are often impossible to serialize. With a state machine you can just take your state and write it to disk, but with coroutines that state is hidden in the stack of the coroutine and the stack is not a first-class data structure you can access by normal programming language means.
grumbel··on Meta’s Reality Labs prototype hardware
The only problem with Cardboard was that Google didn't it follow it up quickly with a real standalone VR headset, nor did anybody else for that matter. The Oculus Go only came out four years after Cardboard. And it wasn't until Quest1, a whole five years later, that we got 6DOF tracking in a standalone headset (Google's own Lenovo Mirage Solo can do it too, but that feature never made it out of Beta).

The thing that shouldn't have existed was Google Daydream. It brought no real advantage over Cardboard, yet was a whole new completely incompatible VR platform from Google, incompatible in both direction no less. Why they couldn't just implement the additional features on top of Cardboard I'll never understand.

Ironically, Cardboard is still alive and you can find some apps that still support it (e.g. Youtube). While Daydream is dead and can't even be used on modern phones.

That aside, content was a huge problem for Cardboard and other 3DOF headsets. VR video still looks great on them, but it took years before there ever were any real consumer VR cameras. And a whole lot of time was wasted with VR360-2D content, which looks utterly mediocre in VR, while good looking VR180-3D got very little attention for some reason.

grumbel··on Meta’s Reality Labs prototype hardware
Meta is targeting one billion people in VR. They have stated that numerous times throughout the years. Meta does not like gaming, they tried numerous times to move away from it and focus on social VR. At their conferences over the last few years gaming gets a quick trailer show and than they go back talking about the Metaverse.

Meta's focus on games right now is not by choice or by vision, but simply because that's the only VR product they have that actually sells. Kids love Quest and VR. Nobody bought the QuestPro for social VR or for work-in-VR. Horizon World doesn't seem to gain traction either, everybody is still doing VRChat. Gaming was also not their idea, the gaming focus came from the original Oculus company they bought, they just have been unable to move past that.

The difference between Apple and Meta is that Apple has a clear vision, they are building a device that works as full desktop replacement. They completely skipped any of the classic VR topics at their presentation. VisionPro is "just" a better monitor, everything else is optional and come later.

Meta on the other side doesn't really know what they want, they have this vague vision of VR being big in the future and they want to secure their place in that future by throwing money at it. But they don't know how to get there. So they have a lot of gaming here, a bit social there and a bit of VR-as-desktop, but none of it is polished, lots of it isn't even functional and nothing is ever used to its full potential. Worse yet, they handicap themselves, their VR must be mobile, since they want to own the VR platform, so their PC support ends up lackluster, despite all the best VR content still being on PC. They want to have ads and collect data, so kids under 13 aren't allowed on their platform, despite being very popular in that age group. It's all a mess.

grumbel··on Meta’s Reality Labs prototype hardware
> Meta has no real interesting in delivering anything other than a product they can ship to millions of people on day 1 at a low price point

Ahm, are we forgetting the QuestPro? Meta tried to build their "Apple headset", 5x the price of a Quest2 and loaded with all the R&D features they could find. And it was a colossal flop. The whole thing is getting shutdown little over six months after release.

They forgot to include critical features like a depth sensor, they couldn't figure out what to do with the eye tracking in their software, face tracking wasn't a feature anybody wanted and reduced the battery dramatically when actually used, passthough cameras are a blurry mess, resolution is too low to work as actual desktop replacement, etc.

The whole thing, despite 5x the price, was firmly stuck at just being a slightly better Quest2. Meta spend the last seven years building a portable Oculus Rift, and they succeeded at that, but despite billions in R&D they haven't really managed to advanced VR beyond that point.

grumbel··on I'm betting on HTML
I would care more about markup if the browser would actually be able to do anything interesting with it, but for most part markup is a just a default style for an element, the semantic meaning is ignored. The <time> tag is about the last one I can think of that actually made a difference from the users perspective.

There is also a lot of markup for really common tasks still missing, I'd like to see an <advertisment> and something to handle pagination at the browser level (rel="next" has been around for decades, but browser don't care). And more broadly, I'd really like to see much better support for long-form documents in HTML or at the very least native ePub support in the browser.

grumbel··on Apple Vision Pro developer kit
I don't think so. While you can artificially stimulate the vestibular system, getting that to work precisely across all the users of the headset would be rather tricky.

But the even bigger issue is that even if you overcome that, you still have the fundamental problem that a correct vestibular response is extremely important to keep users from falling over. If you artificially give people the impression that they are accelerating or decelerating, they'll automatically compensate and lean into it and just fall over. This happens already often enough just from the visual stimulus alone, if you mess with the vestibular system those accidents would get a lot worse.

That said, maybe there is some middle ground that could work where you don't create a real virtual vestibular input, but just a little jolt to notify the brain that something is going in, even if that something isn't matching the visuals. Many people report having less issue with motion sickness when they walk in place or bob their head. PlaystationVR2 has a small rumble motor build into the headsets with is supposed to help a little bit here as well.

grumbel··on Apple Vision Pro developer kit
I think the biggest problem is the controller and locomotion. Almost all VR games available right now use controllers to navigate around the virtual world. Meanwhile Apple has shown nothing that allows you to move around in the virtual world. All locomotion you do on Apple's headset happens in the real world via AR, the fully immersive experiences are basically limited to standing or sitting in place.

This has the advantage of dramatically increasing the comfort, since motion sickness is largely caused by artificial locomotion with the controller, but it also drastically limits what kinds of games you can do in VR.

I would expect them to expand on this area eventually, but I can understand why they wouldn't for the announcement, as for the time being they want the focus on VisionPro as a friendly computer/monitor/TV/cinema replacement.

grumbel··on Why can’t you just roll back from a bad macOS update?
You can do that with NixOS, booting back to an earlier version is literally one click in the boot manager and happens instantly, as the OS is just resetting some symlinks under the hood.

For Debian you might need a file system with snapshot support, e.g. apt-btrfs-snapshot is a package that handles it, though I never used that myself.

← PreviousPage 9 of 22Next →