phh

Alright, there are a lot of stuff in Ready Player One, so I'll formulate more in depth what I meant:

Having one VR universe from which you can join various worlds where a world is pretty much its own game, and you can join friends in that, and it needs to be "alive enough" to feel like actual social interaction. Pretty much a very big MMORPG.

Several games actually have most of that (including Meta Horizon), except that the games inside are very crude, that the transition from one world to another is clunky, and it's obviously completely lacking emotion (which Meta is trying to solve, I don't know whether they will).

There are indeed other stuff in Ready Player One: - walkpads: That's definitely cool, but they already do exist. They are costly, but same in Ready Player One: Only few people have it. But yes movement in VR is a bit of an issue. I sadly don't expect any progress on that matter (give me a trackball under the hand!), until we can connect to our brain (which I think could be possible 5-10 years from now for simple move forward/backward commands?) - stunts: There are a lot of stunts made in game. I don't think there is any reasonable way to implement that realistically. That being said, stuff like EchoVR are already very immersive imo. - haptic gloves: Okay everyone has it in Ready Player One, and even though we have some stuff, it's nowhere near mass-market. So yes I'll agree this is missing compared to Ready Player One.

Right, I didn't want to expand too much on that matter as I usually make unreadable comments. Most of the reasons I can think of that would prevent me from using headset for work are largely personal (though I think many people might identify themselves in it):

I need to play with stuff with my hands (and I mean random stuff, not just always the same toy. Like right now I'm busying my hands with mounting/dismounting some lego that requires putting a bit of force in it, not really something you can do with a controller). I need to look at my smartphone. I might do interruptible stuff which would require corner vision (say I'm playing on TV, waiting for my build to finish that I can see in the corner of my eye). I'm chewing through gums (k I could probably do that with headset on)

Possibly all of those stuff could be fixed with better hw and sw. But. BUT. One of the feature I really like about current day VR, is that I am disconnected from everything. Which would pretty much contradict fixing every issue I have mentioned.

An addition after seeing a sibling comment:

It's more annoying to setup, but I also like playing Euro Truck Simulator 2 in VR with a 1080 wheel. As someone who doesn't have a driving license, that makes me see why some people like to drive to clear their mind.

I like it quite a lot for gaming as it's much more immersive than PC gaming (I don't have any other gaming console). I play both wire-free on Oculus Quest 2 itself, or (still wire-free) from a Windows VM with a RTX3070 pcie passthrough (so I can dynamically switch between ML and gaming usage).

Immersive-ness doesn't only come from just the screen, but also from the fact that you can't multi-task (one could say this is a sad state of my attention).

I play various kind of games. My main usage is "sport", with synthriders and beat saber. I love story telling/puzzle games, though they are quite costly (like 30€ for 4 hours of game play) so that doesn't account for a lot of play time. But I also like a small Cities Skyline (Cities: VR by the people from Cities Skyline), Deisim where you play a god, and Derail Valley, where you drive a train in a miniature world.

As a software engineer I have tried using it for other stuff than gaming, but even though it feels cool, I never found it actually useful. I can keep the headset 4 hours (and probably the whole day) without any issue (assuming no frame drop, which do happen when running from Windows, not in the Quest environment), so that's not the issue. Maybe with a better display resolution it could be better, but honestly I doubt it.

Just a word on the metaverse: I do think it can be pretty cool for socializing (Ready Player One is IMO not that far-fetched), though Meta's universe development still looks extremely primitive compared to everything else that exists. At this point, I think that a Meta Quest 3 with face emotion detection (and improved SoC and display compared to Quest 2), and let competing companies provide a good environment based on that hardware+OS (for instance Rec Room) would be pretty good, but I doubt that could follow Meta's bottom line.

I hope that this is new to no-one generating JSON using LLM, because it felt like the first thing you'd do when I implemented that kind of stuff. That being said, it's nice to have that as a library ready-to-go.

It's interesting because that's my biggest gripe against ChatGPT, it's too verbose and very hard to skim through. Compared to stack overflow answers it'll be easily four times longer. Sure compared to for instance news website, it is less verbose, but those are easy to skim. I'm still able to read through well written text without skimming, though yes most of the texts I'll read I'll skim through, then go back and then to improve my global understanding, which I think is beneficial

Yup I have the same meaning as you. (I'm in north half of France in case someone says they don't)

> The thing that needs more discussion is the closed source stuff Android OEMs / ODMs run at higher ARM privileges rendering all of Google's grand ceremony around Android security moot (other than on Pixels).

You're mentioning security, which is interesting, because here it's not a question of security (don't get me wrong, izat did have security flaws in the past, and I wouldn't bet that modern devices are all properly corrected), but only to who you are sending your private data to.

It's also fun that you mention Google doing security around Android... because the vast majority of people do send their private data to Google!

My personal take is that if we want to control who we send our data to, we need to start from the easy steps: I personally use a modified Android to send the least amount of data to Google. Disabling gps xtra is pretty easy. I think I already have it disabled, but I'll check.

> This seems like much bigger news than it's being received as.

This has been widely known for more than a decade (Sorry I can't provide with a source. I was expecting Google search-in-the-past feature to work, but it doesn't).

As to whether the privacy policy has been declared properly or not, it depends on the OEM: See for instance https://www.bsr.at/mediafiles/Handbuch/CipherLab/User_Guide_... which shows Qualcomm's IZat default integration (This documentation shows an Android 7, but that screen has existed since at least Android 4.4)

> Sure, other chip makers do sketchy things, but is that really where we're at in 2023?

Literally all chip makers literally do that exact thing (except maybe the Unique ID part), just because that's how you're supposed to do A-GPS. For A-GPS you need at some point an oracle that is capable to give you an approximate location. There are some privacy-compatible ways to do that, but I'm not aware of anyone who even looked at it. (Looks like GrapheneOS skims that feature altogether, that's a way that works)

> It's a horrible time too because the embedded 5G chips are about to be part of everything, sending telemetry back

Even though Qualcomm can make things worse, the vast majority of IoS already do just that, and don't require 5G.

Of course I overall agree with you on lost ownership of modern electronic devices (and as you point out, everything became electronic devices), I'm just saying that it's nothing new. I think you need to get back at least a decade to see the unstoppable start of it (though DRMs are much older than that)

> I can only hope the RISC-V combined with people willing to care can lead to a return to a time when people actually own stuff and ownership is something we respect.

What does the ISA change in that regards? I would guess that the currently most deployed RISC-V device are nVidia graphic cards. And you're probably very aware that you don't fully own nVidia graphic cards. I'm also aware of other RISC-V devices in the field that are even more locked down.

> The one issue with using GrapheneOS's connectivity check is that you're broadcasting to the network that you're someone of interest. An Android phone connecting to Google isn't great for privacy but it is normal. An Android phone connecting to a GrapheneOS domain isn't.

Thanks that's an interesting thought.

I had similar thoughts, but going the other way around. I was wondering whether you gave more entropy be not using Google's generate_204, than by using it (= do more IPs do generate_204 calls). But after trying to compute some estimates, I ended up thinking that not sending generate_204 was still better than sending it.

But yes, as you point out, the entropy provided by pinging /another/ generate_204 is much much higher. At this point, it depends if you want to lower the information you send to Google, or your carrier.

PS: Just in case, Android's connectivity check pings a "generate_204" endpoints that as mentioned literally just responds with a 204

Ah thanks, that indeed clears up my confusion.

I don't understand how is that relevant to savings or Apple Card though? I mean sure Apple Card is currently a credit card, but I don't think that's relevant? In France most banks just give the customer the choice between credit and debit cards, and we just mostly prefer debit cards. Apple could probably just do that as well.

That being said, I guess the reason there is no Apple Card in Europe, is that you can't make 2%+ markup on card payments.

> Less than half of French people even have a single credit card

Got a source ?

https://www.cartes-bancaires.com/cb/chiffres/ says 76M cards over a population of 67M (50M for people over 18yo). Sure some people have multiple cards, but that sounds hard to believe. (Also that's not counting non-CB credit cards. Those are pretty rare in France, but I have one)

(Maybe I'm misunderstanding what you said in "a single credit card")

I'm a hobbyist in the field, but I concur and fell into that trap.

I wanted to index my photos with some CLIP, and my complexity mind said that I wanted an under-linear database, because that's what databases are for. I took a bit of time around spotify's annoy for that. Then the first results were meh, and I was like "oh frack how do I know whether the issue is in the approximate NN or the CLIP model?". Then I realized I was worried about linear complexity for a 5MB database, laughed, rewrote into a dumb for loop in a jiffy, and could conclude within the hour.

Someone mean would say that this is not a bug, but a feature for the people who are paying for Chrome.

Most modern (less than 5 years old) Android smartphones support Project Treble/GSI, which largely reduce the maintenance cost.

But that Treble technology is actually being used by ubports which does Ubuntu Touch using Android drivers, and porting Ubuntu Touch on any modern Android smartphones is quite easy nowadays (assuming you have kernel source code available). It pretty much just requires rebuilding kernel source code with different parameters and that's it.

People already largely answered on many points, but I'll just answer on one point:

> unsecure

How is security defined? What is your threat model? If your threat model includes includes running random native binary blobs from random places like Android or iOS does, then yes, you can definitely consider it unsecure.

But that's NOT the target paradigm here, and if you consider it is, you're throwing away 90%+ of the GNU/Linux distributions out there. Last time I checked, running GNU/Linux on servers or desktops wasn't considered bad practice, quite the opposite.

Instead, it may be more secure than iOS, depending on your usage, because the paradigm is to install only opensource software that you actually trust.

The issue is largely around available RAM capacity, and sadly RAM price have barely decreased in the last 5-10 years

I don't believe that in 2023 something without:

- a SMS app - a Maps app or API - a location API - a web browser - an alarm app - Remote notification system - app store (and there are many many other examples)

can be called a smartphone OS. Yet AOSP doesn't contain (Google/)Android version of those.

So yes, I dare say AOSP-based OS can be non-Android.

IMO Europeans have much less "we MUST scale to infinity and beyond" mentality than SV culture. Once you have a working product that pays 10 people, just being able to pay those 10 people + a little margin can be enough. And not trying to scale to infinity means cash is used more efficiently, thus small ad revenu can be enough.

If I was making an OS in Europe, I definitely wouldn't mind shipping Google's gmail + youtube Android apps in a non-Google OS. I have very little doubt that the EU's anti-trust would side with me.

Whatsapp does have a PWA variant though, used in India for KaiOS (which is pretty much FirefoxOS)

> My take on why a third mobile OS is hard is that phones don't have a standard BIOS type system that gets all basic I/O working on it. So having an OS that can install on existing phones is a bunch of work.

You can have AOSP-based OS (nowadays "smartphone OS" covers a whole bunch of stuff that isn't in AOSP so I don't consider that it's the same OS) that re-uses original Android OEM drivers. You can take pretty much any Android smartphone (assuming it doesn't prevent you from running your own software) and run an AOSP-based OS on it with literally 0 effort. That's called Project Treble, or GSI. You'll find dozens of ROMs (both Google Android, and Google-less) that will work on pretty much any modern (less than 4 yr old) Android smartphone.

Even better, it is possible to use those drivers meant for Android on GNU/Linux through Halium. There are even tech demos where you don't need original kernel source or anything, you just flash it, and it mostly works on any modern-ish Android smartphone that lets you flash your own software.

Really, the drivers and hardware support has no reason to be a blocker in the 2020s (I'm not saying it's totally free though) to make a new mobile OS, except that you're stuck to support a wide range of rather old vendor Linux kernels.

PS: The future is brighter, and in few years you'll probably be able to make a really brand new OS, that will run in a VM with drivers running in another VM alongside yours

As far as I know the backend to have the driver running on macos is already merged in mesa, so it should already work as-is?

FWIW it is actually worse than it looks like. Even when Xiaomi did/do provide some kernel sources, it's never anywhere near the production code you're running on your smartphone!

It's hard to believe, but the "GPL release" they made were actually full clean rewrite of the hardware support made on top of SoC vendor's (say Qualcomm) kernel release. I have a very hard time understanding why they did, but there were definitely inconsistencies that are very hard to explain another way.

> It is important to realise, that most cultures do not believe in Intellectual Property the eay 'the West' does.

That's definitely true, but it's not really in a "let's take it more a lenient way" kind, because otherwise Xiaomi would have published those kernel sources

You don't need kernel sources to change Android software though.

But I hate Unihertz for other reason. You can pretty much buy their hardware exclusively on their website which is a horrible e-merchant: Their delivery failed (maybe they never sent it, maybe they sent it with the cheapest post service possible, i don't know), and well pretty much they stole money from me. That was a "kickstarter" [1], so I couldn't ask my credit card for a refund.

And that's after I was already a sold customer since I already ordered one before, AND I gave them visibility by having custom ROM support for their devices.

[1] the quotes are here because the device was Google-certified months before, so clearly the product didn't need a "kickstart", they hide behind it to steal from users

> I should have explained better. My sense is that right now there is an implicit expectation in the market that the 250k threshold is no longer relevant, and that the FDIC is implicitly insuring everything that looks remotely like a deposit, regardless of the amount.

If rich people think that, I think they deserve to be taxed at more than 80%, because that's stupid.

FDIC move was perfectly rational and cost-optimized. Had they not done that, there would have been more bank runs, and they would have to actually pay out 250k/account on many more accounts than just SVB's, while ""saving"" SVB cost almost nothing.

> Guess they weren't as independent as they thought they were.

Simply sharing the same name could be dependent enough to trigger bank run...

> The mere existence of Widevine is a mystery for me. How sensible for Netflix is to invest into DRM at all?

Well that investment is pretty low since Widevine is quite easy to integrate and the licensing cost is 0 for Netflix.

The thing is, you do need some Digital Rights Management somewhere. If you just have clear-text mp4 with no auth, you can just share that mp4 url, and you could watch movies directly off Netflix servers (so it would cost even more than torrenting).

You could include temporary tokens in your URLs, but then you need your CDN to be a bit dynamic.

Or you could have a completely static CDN, but the files are encrypted, and you download the encryption key off a server dedicated to DRM that will check you're properly authed. And well, here you've re-created Widevine. Widevine do provide more protections than just that, but that's just free for Netflix, they literally just have to switch some booleans.

> They seem to make user experience worse for nothing.

Yeah I mostly agree. As I mentioned before, there are some usages of DRM that are not completely non-sense. What kills me is the HDCP requirement. Many TVs have HDCP compatibility issues (for instance my Samsung TV has the two 4k60p yuv444 hdcp 1 ports and one 4k60p yuv420 hdcp2 port, so I need to chose between 4k content and true 4k resolution), while HDCP2 is utterly broken: you can find decrypting dongles for 20€ directly from Amazon.

> IIRC you can only play 4k Netflix on a smart TV which controls the entire stack from network to pixels

Netflix won't allow TV makers to make their own Netflix app, they must use Netflix proprietary code as-is, and it's Netflix code that download chunks and forwards them to playback, while TV handles the secure video decoding, so TV can't "control the entire stack"

Also, tv boxes are allowed to playback 4k Netflix just fine without controlling the screen.

All the Netflix shows you can download on the internet do include the 4k versions.

Maybe they'll obsolete older 4k devices that are deemed less protected (though I could probably point them "a few" security flaws to Netflix-endorsed 4k devices so that every is on par), but I doubt it. I don't think anyone ever done that (720p has always been fine without HDCP, 1080p with only HDCP1, before that macrovision was non-breaking) [1], and I doubt Netflix would want to push that badly towards e-waste.

[1] There has been few cases of some devices being revoked, like Nexus 6, but it was usually long after their shelf life anyway