I don’t think it can. Nuclear is cheap at scale, not when you build a single prototype every 20 years. You need large countries with large programmes to be successful.
It doesn’t take 20 years to build. Are we in the west gonna wait for the Chinese to build out Thorium Reactors, are we also gonna wait for high-speed rail to be built out by the Chinese too if it wasn’t for the war with Russia and the Ukraine, the Chinese probably would have built out high-speed rail, all all the way to Europe, by the way, they’ve already built a rail system all the way to Tehran.
And the Chinese have built out a high speed rail system all the way to Ürümqi, China just east of Kazakhstan in western China. (They aren’t messing around).
There was a very interesting podcast that went into all the details of the AI supply chain shortage [1]
The key takeaway for smartphones isn't so much that iphones will cost $150-200 more, which apple customers have shown they can stomach. But that cheap $200 chinese smartphones will need have to hike prices by about the same amount, which will decimate that market.
Forget hyper-optimizing, how about we just stop throwing away gigabytes for no reward? If we get there, then we can start thinking about maybe actually optimizing anything
browser tabs are the perfect examples, period. These things were being done with native widget toolkits twenty years go. Or going back even further, TUI forms. Rich text with straightforward inline graphics is nice, but we're paying an extremely heavy price for it.
Personally I might not even mind so much about the memory except the latency from the garbage collection, swapping, etc absolutely kills interactive performance. My TI-85 had a snappier interactive experience than my "modern" phone.
I think a large part of the problem, at least on the embedded side, is that people really just don't have standards for these things should feel. When I tried out the first gen Nest in a store, my first thought was it was a joke and who would actually buy it with 100ms+ jittery UI lag. But lo and behold, people did.
I would personally love it if efficiency actually became a priority again, but I don't see it happening. Instead we'll get an even more K shaped economy where people with enough money absorb the extra cost without much care, but everyone else will be materially worse off with a slower more limited phone.
Apple has a long history of not caring about budget-concious consumers. I doubt the wealthy decision makers at Apple and other companies are suddenly going to start caring about them now.
Moore's lesser known cousin predicted this as Leslie's Law. Accounting for cost, Moore's law has practically jumped backwards in time over 15 years in the past six months.
Why does it need to be Android? Make a phone that will send text messages and run a web browser and all the people who just need a phone will buy it to save $200.
Also, when software is poorly optimized that implies it is possible to optimize it.
Web browsing is the most resource intensive task given how bloated modern websites are. It’s easer to imagine a cheap phone with a set of optimised native apps but the problem is tech giants are not interested in reducing footprint of their apps at the same time actively hostile to 3rd party clients.
> given how bloated modern websites are. It’s easer to imagine a cheap phone with a set of optimised native apps
The age old truism in tech is that it's easier to hyper-optimize a single chokepoint (i.e. the platform) than trust distributed actors to each optimize their own thing (i.e. app developers).
Because optimization takes money. Sometimes lots. And distributed non-platform actors have their own priorities like features.
That's why you get enormous success stories like x86, TCP/IP, HTTPx, and Javascript engines providing increased performance, but the "developers as a whole self-optimize" dream remains a perpetual mirage.
(Outside of gaming console hardware, and even there arguably true before middleware)
As already pointed out, your awareness is very geographically limited. In my country (and in a few others I’m familiar with), what I most need a smartphone for is as the obligatory second factor for strong authentication to all kinds of services both governmental and private. That’s done through a bank app that only runs on Android. Elsewhere, people might find that their local public transportation or similar things can only be paid using an app.
> Why does it need to be Android? Make a phone that will send text messages and run a web browser
I think we're quite a ways past the point where most people would be satisfied with phones that only do voice, text, and web browsing. People have become quite accustomed to phones that can run arbitrary applications and games. There's quite a sizable population for whom phones are their only computing device.
There is nothing stopping it from running arbitrary applications, those applications just wouldn't be Android apps. Which is just as well because 99% of Android apps are just a bloated skin over a web page anyway.
I can't actually think of the last time that I wanted to (rather than was forced to) use a non-messaging non-browser app on my phone.
Because you need these apps to do almost anything.
I went to a music festival last weekend, ticket has to be on the ticketmaster app which is android or iOS, the official app for the festival has timings and updates for the event which are very important else you will miss stuff that is announced, even safety warnings.
My train tickets were on the app, I needed it to book the Uber as well.
20% off drinks and food if you use the paypal app. You just can't do things without Android or iOS.
Thousands of people parked in random fields, parked in daylight, need to find car at night = Airtag/ map.
Everything is an app these days. It's a lot harder to do everything without such things. You wouldn't have been able to attend without such a device.
For a start, most of the world does not use text messages. They use WhatsApp. (Apart from a few countries where WeChat, Telegram or Line are more popular.) As far as I know the US is the only country that still uses SMS/RCS.
I don't use WhatsApp, I use Signal, Messenger, iMessage and Snapchat (yes I know but these things have inertia). I'm not exactly alone here. You'd need to write third-party clients for all those apps as well.
I was about to say this already exists, but Meta discontinued the KaiOS client a year ago.
For all the EU does right, it's amazing that a reviled United States conglomerate owns their entire social sphere. I hope they can change that and show the world a model of how to do so, especially given how high the stakes are.
Unfortunately that EU WhatsApp interoperation thing isn't really what we wanted. I assumed it would mean that you would be able to communicate with WhatsApp contacts using a non-WhatsApp app (kind of like Pidgin back in the day).
Meta obviously don't want that and they've done a sneaky thing to make it useless: if you connect a third party app to WhatsApp, from that third party app you can only chat with other WhatsApp contacts that have also connected the same third party app to WhatsApp. So if you write your own WhatsApp client that runs on this low power phone it will be completely useless because to chat with your friends they would all have to manually connect their WhatsApps to your client, which of course they won't do.
That’s pretty screwed up. So even if I setup e.g. Telegram as an allowed app in my WhatsApp account, I can’t communicate with friends on Telegram because they haven’t (why would they, they don’t even have WhatsApp) configured their WhatsApp to use Telegram?
>The Alchian–Allen effect was described in 1964 by Armen Alchian and William R Allen in the book University Economics (now called Exchange and Production[1]). It states that when the prices of two substitute goods, such as high and low grades of the same product, are both increased by a fixed per-unit amount such as a transportation cost or a lump-sum tax, consumption will shift toward the higher-grade product. This is because the added per-unit amount decreases the relative price of the higher-grade product.
I'd expect the iPhone price hikes to be less even in absolute $ than with the cheap phones. iPhones already had relatively large margins (as a %) for the newly increased costs to partially eat into.
Why would you expect Apple to give up their margin? The DRAM price hikes affect every smartphone vendor, even Samsung through opportunity cost. That means the competition will also need to hike their prices. And the competition has lower margin, so they have less choice here.
Because Apple's main competition is themselves a few years ago, when they sold an iPhone or whatever to a happy customer who might return for an upgrade even though it still works. Not many people are switching platforms nowadays.
That's if iPhone costs go up and there are no other changes. As the parent said, competitors' products are also going up in price. I just don't consider the products interchangeable enough. Apple has its margins for a reason.
Also even if they had pure competitors, demand for high-end smartphones is pretty elastic. People will really just buy fewer phones if they get more expensive.
Apple has services and the App Store that still collects 30%. I don’t think they will raise iPhones prices by much but rumors say the iPhone Foldable is coming and that will be $2K+. People will pay it and that will subsidize the other models.
There's a limit to what even a wealthy customer can stomach. They need to consider what costs them more, Apple eating $50-$100 of the cost, or people holding on to an iPhone one year longer than usual.
These announcements have another effect of boosting sales now. Summer is usually a sales slump so selling more now is probably good for Apple.
The schedule is predictable now but in case you forget, you can always tell when the new iPhone is right around the corner because AT&T starts spamming your email and SMS with 'reminders' that you can get an iPhone 17 right now! Or for Father's Day! Or because you're eligible for an upgrade! Or...!
Gotta get rid of as much of that stock as possible, but keep collecting 16s to harvest for repair (not that I object to improving repairability ,per se)!
We'll see what happens but I wouldn't be shocked or offended if China obliges the state-sponsored memory makers to support other local industry, beyond AI. The consumer electronics world shutting down would be immensely bad, for China, and for many: someone steering us all away from this path seems highly advisable for everyone.
> But that cheap $200 chinese smartphones will need have to hike prices by about the same amount, which will decimate that market.
Hmm. Even if iPhone users can theoretically stomach the increase, they have many other options available, whereas if the cheap $200 phones are the bottom of the market, there's no other real options.
I'm in the ~$450 USD Pixel range atm, and never buy the current flagship or anything. If that increases by $200, I'll look to the used market for the same phone. I really don't care that much about it, and it mostly acts as a fancy 5g modem for tethering. Plenty of younger people are already reverting to more primitive phones or physical media, and I wonder if it's just older addicted richer millenials that'll keep buying at even more than the already idiotic prices.
> Plenty of younger people are already reverting to more primitive phones
That's an effect that has always been claimed (younger generations rejecting new tech and going offline/low-tech/anti-...) but it's never been more than a minor very temporary fad. In the mid to long term, younger people are always at the forefront of tech adoption and it would be very surprising if it was different this time.
Frankly I don't think I've ever heard of this being claimed for prior generations at a scale that any other generation would notice. There was always that kid who kept their CD collection or their dumbphone longer than most, but it was more of a niche thing to do. That said, I don't wholly disagree that it's a fad, time will tell.
Considering there hasn't really been (to my knowledge) other instances where involvement of media and technology has been remotely as intense as now from day one in a kid's life, it doesn't seem like any prior time period would be as likely to produce a sense of repulsion as now. Sure, I didn't carry my Gen X relative's obsession with television into my adult life, but something like that I'd consider to be a bit of a tenuous comparison to make to the breadth of insidious mind-altering media delivery conduits we're constantly exposed to now.
It's about what role tech plays in life. The claim that young people are increasingly detoxing and embracing offline life is just not true. More than any earlier generation, being online to stay in touch with friends and the general happenings of the world is integral to most young folks life. The same thing will happen with AI chats.
My grandma rarely uses google in her life. My mom uses it rather often. To my generation it's the portal to knowledge background for most of what I do. (Substitute any search engine for google if you must.) Similar with instagram, instant messanging etc. AI chatbots will go that way too and so will the next trend.
The point was that if there was a time when people who'd otherwise be buying the latest smartphones for relatively mundane purposes—spending arbitrary amounts of money for them—would balk at the idea, that time could be now. Nothing to do with chatbots.
Money is tight, jobs are scarce; for many people on their way into college or even in the middle of what would be their career, there isn't a clear answer to the question of whether the current tech landscape is anything but a threat to one's future or their mental health.
Seems like kind of an anecdote vs anecdote situation where I see people cutting back severely on their exposure and being more frugal or retro, and you don't, which is fine, but I'm not doing anything more than speculating at hypothetical future behavior based on real world conditions and what I've seen.
There's simply not much required from a hardware standpoint to facilitate multi-modal communication patterns if someone's practicing a limited-exposure relationship with tech.
For the average consumer, I'd argue that any phone made in the past 10+ years would be absolutely fine.
It issues are: battery life/battery replacement, lack of updates, developers targeting newer devices only.
In Apples case a good solution would be to rollback at least part of the liquid glass UI updates, as it severely affects older devices. Then announce upfront that every Apple phone will be supported for no less than 10 years after the device was removed from the market. That would be good for everyone, except Apple shareholders.
I understand that Apple pricing, compared to inflation haven't changed that much since the first iPhone, but for many of us it really does push the limit. I simply do not get enough value to justify purchasing a new iPhone, or in many cases a second hand one. My perceived value peaked around the iPhone 7 era, everything after is pointless. Apple doesn't really cater to my needs, and that's their choice, I just feel a bit stuck.
I do use an iPhone 6S. The hardware is fine, but I basically can't install apps because they all target newer OSes that can't be installed on it, and the older versions of those apps available are broken.
I think that's the unfortunate cycle of iOS and iPadOS. Their hardware more or less served it's purpose eons ago and there's little reason why it couldn't still, aside from OS compatibility. Even as a mac user, I've only ever purchased on iOS product, and that was the iPad 3. No subsequent iPad seems to have introduced anything more compelling than what the iPad 3 was capable of (which wasn't much), so as the software compatibility story has degraded over time, it still sits here serving as an occasional epub reader. It would feel bizarre to spend thousands on a new one to effectively be able to do what I did with it in ~2012.
My android phones eventually run their course due to excessive physical damage, theft, etc.. but since I've never paid much for them, it's usually not so disruptive to just pay a few hundred and get something used or basic to replace it, or order a new screen and repair it.
Not enough to have stopped the iPhone 17 Pro line from being a runaway success that even Apple—famously excellent at projecting demand and already invested in selling as many as possible given its the flagship model line of their flagship product line—completely underestimated the demand at launch.
I’d love to know how much “plenty” in the parent’s perspective stacks up against just this one individual model line and whether it is at all distinguishable from noise.
Surprising amount of hostility to what was a speculative anecdotal observation. I'd be surprised if many people under the age of 17 would be buying a $2000+ phone anyway, but given tough economic situations, decreasing opportunity, and decreasing stability, the likelihood seems low that significantly increased prices for effectively the same (boring) product would be absorbed by anyone other than the richest or older established crowd. Whether it holds true at global level, or in your social circle, I don't know.
If the smartphone is not economically viable, it will go away
Apple’s margin targets aside, the prices are rational
Trendy teens and 20 something’s still have iPhones, many just also have point and shoot cameras. This is more of a desire to be present in some contexts alongside aesthetics (of the photos and the gear), than a rejection of having 2 teraflops in their pocket.
Apple sure is lucky that the supply chains for those low end Chinese phones are in a country that is notoriously bad at building new manufacturing capacity and prioritizing their own industry </s>
I enjoy windows 10 hugely now that it is out of support. It became way better when microsoft started tormenting the users of win11 instead of win10, and now that windows update doesn't bring new catastrophes and unexpected reboot, the OS is finally not interfering with usage anymore.
> It became way better when microsoft started tormenting the users of win11 instead of win10, and now that windows update doesn't bring new catastrophes and unexpected reboot, the OS is finally not interfering with usage anymore.
Yesterday when I booted my windows 10 desktop PC I got a bunch of popups (Win32 MessageBox) about errors in some O365 AI dll files.
Turns out some MS AI software was silently installed on my PC in late may.
I do not have MS Office or anything that should require any AI software.
What you probably have is a component of O365 as an updater installed - check your programs and features because msft bundled this awhile back and once its removed it will stop the update treadmill.
Didn't they create Windows Installer (i.e. MSI) originally to meet the needs of Microsoft Office? And now they have MSIX to replace it, both of them built into Windows, but they still need a third installer system for Office, which is now also apparently built into Windows?
You should try Windows 7, then. Boots in no time at all, feels EXTREMELY snappy, and nowadays, there are UEFI loaders and a thing called VxKex-NEXT which adds the few new API calls Windows 10 introduced and thus makes W7 run many W10-only apps.
VxKex/VxKex-NEXT really is an amazing bit of work. It helps me run the latest Apache, PHP, Go-based tools (ngrok/stripe cli), node. It’s been a game changer for keeping W7 going.
Note, there is a way to turn on extended support (updates). I'm getting updates on my w10. And random restarts, argh. Googling it should be enough to find it.
Windows 10 wouldn't let me run Minecraft because the OS was out of support. I upgraded to 11, and it wouldn't let me run without logging into the OS Store, in addition to Minecraft. I removed my Windows partition completely. I run MC on Linux now.
The computer is not exposed to the WAN (behind a firewall), the main way it could get infected is via a vulnerability in a browser, but these do get updated. And OS updates don't really protect you from malware in executables you install anyway.
The other potentially obvious question is why bother using an OS that’s out of support when Linux is so good?
I left Windows 11. The last straw wasn’t Microsoft accounts or Windows updates. I actually thought the OS was fine, most OS updates actually added great new features, and anything I considered an annoyance was easy to disable permanently.
Toss your Windows 11 ISO into Rufus and disabling things like Microsoft account requirements is a trivial process.
What I actually rage quit Windows over was AMD graphics drivers and a couple of my video games crashing.
What caught me by surprise is just how little I’d miss it. I thought I’d need to dual boot or run a Windows VM for little random things. Nope, I just don’t need them.
I didn’t expect to find an OS with more software that I tend to like better. Like my email client, where I moved from Thunderbird to Evolution and for the most part I find that to be a step up in user experience.
> The other potentially obvious question is why bother using an OS that’s out of support when Linux is so good?
I'm in the process of setting up a Linux desktop to replace my Win10 one, and for me it's these (if anyone has suggestions for migration or replacement, I'd love some opinions!)
- Lightroom. If anyone knows how to either run this under Linux, or migrate an entirely catalog of photos (plus edits) to something open source (including the Negative Lab Pro plugin), that would be amazing.
- MusicBee. There just does not seem to be a good music manager for Linux that can replace MusicBee. I rarely use it as a music player, there are dozens of great options for Linux music players, but MusicBee feeds my Airsonic instance, and I have not found a good way to manage music graphically in a way that maintains this setup.
- Games. This is really getting better and better each year...but I regularly play Microsoft Flight Simulator and haven't even tried to get that running in Linux yet (anyone have good experiences getting this working?)
> - MusicBee. There just does not seem to be a good music manager for Linux that can replace MusicBee. I rarely use it as a music player, there are dozens of great options for Linux music players, but MusicBee feeds my Airsonic instance, and I have not found a good way to manage music graphically in a way that maintains this setup.
For a traditional "all batteries included" collection management music player try Strawberry and Quod Libet. You should also look into the MuiscBrainz Picard tagger, it's a bit unwieldy to use but is very powerful once you learn it's wonky workflow.
I used Quod Libet as my main music player for a year or two, but I found it lacking and iirc sometimes laggy (it is written in Python!). Updates also seem to be infrequent.
It seems that various MS Flight Simulator editions run on Proton. Focus on recent ratings, as the overall score reflects often reflects old versions of Proton.
> - Games. This is really getting better and better each year...but I regularly play Microsoft Flight Simulator and haven't even tried to get that running in Linux yet (anyone have good experiences getting this working?)
It's a major step up in power but the steam deck has really pushed the wine/proton environment to near parity. The only things that really don't work through it reliably is anti-cheat stuff that I really don't want on my machine anyway.
I can't speak for the experience with nvidia drivers but it's pretty amazing how far it's come.
For games, my tip is that if you happen to be in CachyOS or some similar distribution, make sure you use the bore kernel for gaming.
I switched from the standard kernel to the bore kernel and went from a pretty disappointing experience in terms of performance and stutters to a really great one.
For me, I can't switch to Linux because of my accounting software; it's only on macOS. They are very few Linux business accounting software programs suitable for sole proprietors. GNUcash is too hard to set up. Online accounting software is not good, because one is giving ownership of one's financial data to another entity, who could deny access at any moment.
I've had five clients who have lost access to their Microsoft accounts permanently due to insufficient, or old, recovery information. SMS can't be used anymore. I've been thinking about recommending Yubikeys, but when older people don't even want to use password managers because they don't trust them, that's a hard sell.
The biggest problem is Microsoft changes the rules and requires all of these features, but doesn't tell any normal users of the changes nor the addition of the features.
Namely, it's the "blockers" one hasn't found suitable replacements for.
Love Linux, but Nvidia drivers are still shit on it. I'm not willing to take a performance hit for the convenience. Which I guess is a little ironic, given you left Windows over AMD driver issues.
The last time I had instability on a Nvidia card in Windows turned out to be a faulty card I had to RMA.
It is funny isn’t it? I imagine maybe a clean install or some other intervention might have helped me. The instability only affected two specific games at least on a regular basis and it was kind of a new thing.
But yeah, switching to Linux with an AMD card is basically an upgrade compared to Windows.
I would guess because "so good" does not equate with 100% and presumably the user's needs fall in that 5%.
Linux has been usable for non proprietary software for decades now. The fact that people are refusing to jump ship even when Windows actively undermines them and itself speaks volumes of people's aversion (or inability) to switch OSes.
Oddly I’m like mostly using proprietary Windows software on my Linux machine these days (games).
I also think the AI era goes very far in eliminating those 5% problems. I have a mostly non-technical friend who set up an old laptop with Linux for the first time and he told me that he’d never have been able to do it on his own without AI. Anytime there’s an issue, his solution is just a quick question or copy/paste away.
Running Windows 10 Enterprise IoT LTSC still gives you updates until 2031 with the added benefit of no app store. I run it as my main OS since last October and have yet to encounter any issues.
>Running Windows 10 Enterprise IoT LTSC [...] have yet to encounter any issues.
It depends on the type of software a user runs. I installed Windows 10 LTSC on a friend's computer last year thinking she could run it for at least 5 more years and just ignore the newer Windows 11/12/whatever.
But she needed Intuit TurboTax 2025 and it requires Windows 11 and it's a hard requirement. The installer aborts on Windows 10. It's not a soft requirement like Adobe where they only support Windows 11 but their installer still runs on Windows 10. Autodesk Fusion 360 is another example that requires Windows 11.
I'm guessing if there's a future Windows 12, Intuit TurboTax will be aggressive about making it a requirement that forces the issue even though nobody wants to upgrade to it.
I'd say that comes down to the difference between requirements (i.e. will it run at all, does it use features only found in win11) and support, and the developer's decisions around that. I can appreciate not supporting win10 even if it runs as they have a written or implied burden to make sure it keeps operating correctly for the lifespan, and that may include keeping test systems around or handling bugs that turn up in the OS that's getting reduced support itself, or other factors like drivers. Then there's the question of whether people would be willing to pay for a "your mileage may vary" level of support on something commercial.
Only downside I've encountered using W10 IoT LTSC is that I had the temporarily change the currentbuild key (HKLM\SOFTWARE\Microsoft\Windows NT\CurrentVersion) in registry to 19045 to be able to install docker and WSL2.
From definition, a 0-day is not patched in any system because it's not known. But back to your real question.
The biggest attack vectors are the browser, the mail client and direct network access.
I would never use outlook, edge or connect my computer directly without NAT or firewall to the open internet. And would never open a website without a add blocker.
You can count all other known big attacks(on unpatched Windows 7!!!) on one hand.
1) Remote execution via Wifi Stack
2) Remote execution via True Type Fonts
3) 0-Click code execution via USB Stick Icon processing
Windows update instead gives AT LEAST Microsoft a steady remote code execution on your and millions of other computers. It's a really interesting attack target when you go big. Why I should trust M$ to get the security there right?
I use Windows 10 with a relatively obscure firewall software with a per-process/per-service whitelist, and try to not be stupid on the Internet. I also do regular backups. This should cover most of the risk model applicable to me. Has worked so far.
Why would they need to be any more worried about those now than before?
The same holes exists and have existed for some time already. If he was not worried about them before why be worried about them now? And if you're worried about security holes why not be worried about the ones that exist now?
In general I find it funny that some people think that system is "secure" when it's on the latest version. At time t0 version N is considered "secure" then an update is made at t1 with version N1 and suddenly N is no longer secure. But it didn't change... it's the same version it was before.
Fact is a computer system is never going to be 100% secure.
Win10 is still getting updates if you opt in. My dad's computer started hard locking after a recent update - I spend a few hours trying to fix it, and I'm not terrible at debugging win10 issues, but I couldn't find the problem. Had to reset windows. It worked fine after doing that (of course, there was no actual hardware problem, as I had guessed), but I have to reinstall all his tax software since he does his (and my) taxes on it. So the long term reliability of win10 is rather questionable at this point; I suspect vibe coding has infiltrated the windows update process.
Personally I'm on linux full time now, which has its own issues but enshittification by Microsoft is not one of them.
How is that going to help him? "Our models are so inferior they are not deemed a threat unlike anthropic's"?
I think it is either a missile directed at anthropic, as retribution for not giving the DoD what it wants, in which case it is likely to resolve pretty quickly. Or it is a shift of policy toward export restrictions on powerful LLM and then every model will be impacted as they reach the threshold. In which case this could have massive implications of revenues, valuations, and the whole datacenter buildout. And frankly on the location of the white collar workforce if it is indeed a productivity multiplier, all countries reciprocate, and not all countries can match the US LLMs.
And why would the EU allow exports of chip manufacturing equipment if the US then restricts the export of derivatives of those chips to the EU?
There are many people in this administration invested in OpenAI like the Kushner's. They are attacking Anthropic however they can. You will notice a lot of propaganda on social media sites against anthropic. It's very obvious.
the first one to do IPO will win big. With the government pressing Anthropic, OpenAI IPO will vacuum up the funds that otherwise would have went into Anthropic IPO as OpenAI was falling behind.
>and the whole datacenter buildout.
somebody just did a $2T IPO with the idea of datacenters in space. One can wonder what laws/jurisdiction those datacenters will be subject to.
> How is that going to help him? "Our models are so inferior they are not deemed a threat unlike anthropic's"?
Do you honestly think that this - logic and reason - is going to stop anyone from hyping whatever nonsense he comes up with to the moon and back anyway? Right after the SpaceX IPO of all things?
The other thing is what this will do to 1) the valuations of these companies, 2) their potential revenues and therefore the viability of the current datacenter buildout. Looking forward to the reaction of the market on Monday.
To be honest lots of developers think they don’t need to learn machine code. They just need to learn a language which once compiled will produce machine code.
I wonder if a probabilistic compiler would be fine for the people arguing this. One that sometimes produces machine code that does something else, and sometimes produces machine code that is just broken and does nothing useful. From the same source code.
What if your compiler could be fooled by some other developers into spending thousands of dollars, and still not produce the desired machine code in the end?
Understanding assembly/machine code is optional but helpful. The programming language semantics are enough to reason about what the program is doing. Other tools also help, but are optional for learning how to program.
Using an AI, there is no semantic model that can be used to reason through. You're left without any mental model of the proglblem at all.
I've been arguing for years that is isn't optional and treating it like it is is how we ended up with Electron and 400MB JavaScript websites.
When you have no mental model of the machine running your code or what the physical implications of code mean, you fundamentally lack the ability to reason or care about performance. "Works on my machine" is the original vibecoding.
I mean I don't disagree, but there's still a difference in the kind of disconnect you get. The disconnect is harmful in the high-level language case, but it's dangerous and irresponsible with vibe coding/LLMs.
Also, I would argue that a good enough understanding of computer architecture and a mental model of a process' memory layout gets you there, without knowing how to write assembly. That's still a mental model.
LLMs these days seem to have no problem using language semantics to conceptualize what’s happening in a program. This is my favorite use of an LLM, “why is this library doing x” and then it digs through the library itself in my venv to find an answer.
What I've done is ask the LLM a question like "How do I configure EF Core in this particular way?", then when it tells me the answer, I go and look up that function in the EF Core docs and learn by reading the docs. (Which also tells me whether it's correct or hallucinating; one time the LLM told me "You can do X like this" and the documentation said "We don't yet support doing X, but in a future version you'll be able to do it like this"). Here, I'm using the LLM to compensate for the fact that MSDN search is awful and the bits of info you need are scattered across three different articles, none of which link directly to each other.
You internalize the inner workings of all the libraries in your venv? Impressive! My current project’s uv.lock has ~60 packages in it already, reading and comprehending those tens to hundreds of thousands of lines of code must be time consuming.
You’re also just confidently wrong about the model reading the code. It quotes file paths and line numbers and I open and read those files at those line numbers. For me, hallucinations are much more frequent when it references the docs rather than code because docs are more subjective than code.
This is a normal thing I’ve been doing since at least December.
I have to ask — do you actually use LLM coding tools? Your knowledge on this topic seems really out-of-date.
The fundamental architecture of LLMs has not changed, so knowledge on that cannot be out of date.
Do I internalize the inner workings of all the libraries? Not unless it's necessary. Sometimes it is. If I want to read the source code of the function I'm calling, I can just do that, my IDE pulls up the file with one key combination.
I'm perfectly capable of reading source code myself, I don't need a non-deterministic filter in between.
Compilers are deterministic and, luckily, not agentic.
But yes, it's not obvious (or perhaps even likely) that it just happens that current high-level languages are the "correct" optimal level of abstraction at which you can ignore the sausage-making details at the lower levels. Ultimately, of course, it depends on the use case. Something like Python is so far removed from machine instructions that knowing assembly hardly gives the programmer any additional value.
(Also, obligatory reminder that assembly and even numeric machine code are also abstractions, an "API" provided by the CPU. Instructions get split or fused into micro-ops, named registers are a backwards-compatible abstraction over a much larger register file, instructions get reordered and executed in parallel depending on their data dependencies, a large fraction of the total transistor budget is spent on multi-level caches and cache logic to maintain the illusion of fast access to a single, uniform memory space...)
It requires the claim to be made with "willful disregard for the truth". Notifying someone, especially with a cease-and-desist on fancy letterhead, makes it legally clear that they know better, and thereafter would be definitely libelling you (assuming the claims are in fact untrue and harmful). But you can still sue them for the claim prior to the notice, you just have to prove they should have known better prior to making the claim.
Libel/defamation laws vary wildly across different countries. Sometimes true statements can be considered defamation, sometimes not. The same goes for intent.
And conveniently, by making your machine non upgradeable, it allows the manufacturer to enforce market segmentation / charge a huge premium for small RAM upgrade (a la Apple)
LPCAMM2/SOCAMM2 exist, heck I think Framework is using LPCAMM2 in one of their new laptops.
Heck, I'm willing to bet that a lot of manufacturers would rather go that route than soldered in, if for no other reason than the relative cost of warranty work between the two.
However, people probably need to stop being obsessed with ultrathin laptops for that to happen.
> However, people probably need to stop being obsessed with ultrathin laptops for that to happen.
I've never been able to understand this. Once we made it down to ~20 mm (which for the record still accommodates dual-stacked SO-DIMMs, a 2.5 inch bay, and a user replaceable battery but not an RJ45 jack) I don't understand what the practical impact of any further reduction is supposed to be. Regardless of how thin you make it the thing will still be a massive rectangle that you can't flex or press on.
> Regardless of how thin you make it the thing will still be a massive rectangle that you can't flex or press on.
There's very wide variation between laptops in how noticeably they'll flex or yield or creak when pressed. Laptops with a build quality that actually feels solid are far from being ubiquitous or even a majority.
Doubling the thickness of my MacBook Air would probably make it regress on that solid feeling, unless the weight was also significantly increased.
And regardless of whether current laptop form factors could accommodate a 2.5" drive, there's no use in doing so. That drive form factor is entirely obsolete for laptops and is just a waste of space and materials, and has been for about a decade.
I wasn't saying that I want a 2.5 inch drive, I was merely listing off a number of rather large things that fit just fine within a 20 mm budget.
I'm not sure why you seem to think that making something thicker would reduce the stiffness or strength. It's generally the opposite - see the concept of a torsion box. Anyway that wasn't the point. The point was that regardless of how thin you make the thing it will forever remain a cumbersome and delicate item that you have to treat with care when packing so what meaningful positive impact does shaving off those last few mm have? It's never made any sense to me.
I came here to say just this myself! Modern DIMM formats make SFF/portable builds with unified memory pools far more plausible than prior designs. There's absolutely no reason desktop machines couldn't implement similar DIMM formats or design a new board standard around something similar.
Unified memory doesn't have to be soldered on or serviceable. That's a choice Apple made because it fit their product vision, but it's not mandatory in the slightest.
LPCAMM2 is available in real systems at 7467MT/s and 120ns latency, vs apple (and intel) at 9600MT/s (and apple soldered memory at 100ns latency).
I don't know how linear or sensitive CPU and GPU benchmarks are to such a 20% slowdown, but i don't think Apple wants to pay it. And it looks like the next generation will be even closer to the SOC.
how about the LPCAMM route? Framework uses LPCAMM2 in 13 Pro laptop mainboards and claims that it satisfies the iGPU and NPU hardware without needing soldered RAM
Thought getting sleep right was something that happened before MS decided they need to be able to wake your PC any time they want and not hardware related much.
I mean is it possible to make unified memory systems with good performance or is it not really feasible due to memory timing/trace length issues?
It’s possible if you’re willing to go with much slower RAM than GPUs like but CPUs often use. Thats what integrated graphics laptops have done for a long time right?
But can you get high end CPU and GPU performance with unified memory and maintain user upgradable memory in a reasonable way? Thats what I don’t know.
> I mean is it possible to make unified memory systems with good performance or is it not really feasible due to memory timing/trace length issues?
LPCAMM and similar solutions exist, but have never been demonstrated running at speeds that match what the leading soldered memory systems are using; there's always been some speed penalty. I'm not sure we've ever seen a system demonstrated using LPCAMM or similar for a 512-bit bus to match Apple's Max tier SoCs, so it's somewhat of an open question whether those solutions can offer upgradability at the high end of the market for unified memory systems.
> LPCAMM and similar solutions exist, but have never been demonstrated running at speeds that match what the leading soldered memory systems are using; there's always been some speed penalty.
LPCAMM2 supports up to 9600MT/s, which appears to be the same speed Apple is using.
> I'm not sure we've ever seen a system demonstrated using LPCAMM or similar for a 512-bit bus
Servers commonly use a 768-bit DDR5 memory bus per socket even without LPCAMM and LPCAMM allows shorter traces than traditional DIMMs. It's basically down to most existing DDR5 system boards/sockets having been designed before anyone was trying to run LLMs on consumer hardware, e.g. AM5 has a 128-bit memory bus and you're not changing that without a new socket. But every memory generation gets a new socket anyway, and the existing Threadripper Pro socket has a 512-bit memory bus as well.
Moreover, making the bus wider is "easy" -- the main problem with it is that it adds cost. Apple's least expensive machines use the same 128-bit memory bus as most PCs and the ones with the 512-bit bus cost as much as Threadripper if not more.
> LPCAMM2 supports up to 9600MT/s, which appears to be the same speed Apple is using.
The difference here is in what the standard defines on paper vs what is actually shipping in products and readily available off the shelf. Who's selling a whole system with LPCAMM2 certified for 9600MT/s? Intel's current-gen Panther Lake top of the line laptop chips are rated for 9600MT/s when using soldered LPDDR5x but only 7467MT/s when using LPCAMM2, according to their current datasheet: https://www.intel.com/content/www/us/en/content-details/8721...
That puts the current Intel-with-LPCAMM2 supported memory speed at 1.5 years and counting lag behind Apple's shipping memory speeds. Intel's own shipping memory speed moved past 7467MT/s a few months earlier than even Apple's.
> Servers commonly use a 768-bit DDR5 memory bus per socket even without LPCAMM and LPCAMM allows shorter traces than traditional DIMMs.
> Moreover, making the bus wider is "easy"
Citations needed. Servers aren't anywhere close to 9600MT/s yet; Intel and AMD are at 6400MT/s. The trace length advantages offered by LPCAMM2 don't necessarily mean the traces for the sixth or eighth channel would be short enough for 9600MT/s (which again, is not yet available even in a 128-bit configuration in shipping hardware). Adding more channels to even a LPCAMM2 configuration means adding more trace length, because only two modules can actually be adjacent to the CPU socket. (Maybe you could get to 512-bit with modules on the front and back of the board while maintaining trace lengths short enough to reach meaningfully higher speeds than regular DDR5, but so far nobody is doing that or even talking about it.)
> That puts the current Intel-with-LPCAMM2 supported memory speed at 1.5 years and counting lag behind Apple's shipping memory speeds.
It turns out Apple isn't getting 9600MT/s either. I assumed that soldering would be getting them at least what LPCAMM2 is rated for, but if you actually do the math, they're getting ~8500MT/s for their most expensive systems and ~7500MT/s for the others.
> Servers aren't anywhere close to 9600MT/s yet; Intel and AMD are at 6400MT/s.
Servers use conservative timings. EXPO memory kits above 6400MT/s are available for Threadripper with 8 channels. And again, these are using traditional DIMMs with longer traces rather than CAMM, but they're still managing an extremely wide bus with close to the same performance.
> The trace length advantages offered by LPCAMM2 don't necessarily mean the traces for the sixth or eighth channel would be short enough for 9600MT/s
CAMM modules use a compression fitting to attach the chips to the system board using approximately the same amount of space as the solder pads would for soldered chips. If you get to the point of having so many channels that the chips are in the way of the other chips then the soldered ones have the same problem.
> (which again, is not yet available even in a 128-bit configuration in shipping hardware).
A single LPCAMM2 module is a 128-bit bus. Every system that uses it has at least that.
> Maybe you could get to 512-bit with modules on the front and back of the board while maintaining trace lengths short enough to reach meaningfully higher speeds than regular DDR5, but so far nobody is doing that or even talking about it.
Nobody is really using a bus that wide with soldered memory either though, outside of the couple of Macs that start at ~$3500 and are getting the same speed Framework does with LPCAMM2.
> Framework already sells LPCAMM2 at 8533MT/s with full validation:
From your link:
> Framework Laptop 13 Pro (Intel® Core™ Ultra Series 3) supports one slot of LPCAMM2 memory up to 96GB at the native 7467 MT/s speed. It is compatible with LPCAMM2 modules with memory speed rated above 7467 MT/s, but the speed will be capped at 7467 MT/s because of the platform limitation.
The modules in question can only theoretically operate at 8533MT/s. Framework has yet to sell a system where the modules actually operate at more than 7467MT/s.
> It turns out Apple isn't getting 9600MT/s either. I assumed that soldering would be getting them at least what LPCAMM2 is rated for, but if you actually do the math, they're getting ~8500MT/s for their most expensive systems and ~7500MT/s for the others.
You're either doing the math wrong, or just plain looking at the wrong systems. Try looking at the M5 generation.
> CAMM modules use a compression fitting to attach the chips to the system board using approximately the same amount of space as the solder pads would for soldered chips. If you get to the point of having so many channels that the chips are in the way of the other chips then the soldered ones have the same problem.
Yes, that's a problem, and Apple has solved it by moving the DRAM on-package. Datacenter GPUs have also solved it that way by putting the DRAM on a silicon interposer to allow even wider bus widths. Soldering standard DRAM packages on the motherboard is not the limit of how memory can be soldered down.
> A single LPCAMM2 module is a 128-bit bus. Every system that uses it has at least that.
Yes, 128 bits at lower speeds. Did you forget that the whole point I'm making here is that the speeds are not the same?
> Nobody is really using a bus that wide with soldered memory either though, outside of the couple of Macs that start at ~$3500 and are getting the same speed Framework does with LPCAMM2.
The Mac Studio with the M3 Ultra is actually running the DRAM at a lower frequency than what Framework and other Intel-based systems could, but more than making up for it in bus width, to provide far more total memory bandwidth than any plausible LPCAMM2-based system that could be built today.
> You're either doing the math wrong, or just plain looking at the wrong systems. Try looking at the M5 generation.
The M5 generation isn't "1.5 years old" and even those aren't all that speed. The M5 Max with the 32-core GPU is ~7200MT/s, while the one with the 40-core GPU is over $4000.
> Yes, that's a problem, and Apple has solved it by moving the DRAM on-package.
There is no "package" here. Apple's processors are soldered to the logic board, as are Intel's in laptops. The DRAM Apple uses is standard LPDDR5 from the normal OEMs. Have a look at the LPCAMM2 module. It has four standard DRAM chips on the top and a connector on the bottom. DDR5 channels are really 32-bits, so the 128-bit module has four channels, four chips. The module is barely any larger than the chips themselves. It's not saving significant space by soldering them, it's just an alternative means of attaching them to the system board in the same place.
> Yes, 128 bits at lower speeds.
At the same speeds Apple was shipping a few months ago. Apple being the first to ship LPDDR5-9600 when it was that recent doesn't imply that it needs to be soldered, it implies that they're a huge company that can pay for early access to the new thing whether it's soldered or not. 9600MT/s LPCAMM2 modules have already been announced -- it's not a technical problem, it's an "Apple and OpenAI are buying out the fastest DRAM right now" problem.
> The Mac Studio with the M3 Ultra is actually running the DRAM at a lower frequency than what Framework and other Intel-based systems could, but more than making up for it in bus width, to provide far more total memory bandwidth than any plausible LPCAMM2-based system that could be built today.
By this logic the thing to beat it is the 8S Xeon servers from almost a decade ago with 48 channels of DDR4-2666. Or existing 2S servers with 24 channels of DDR5-6400.
Ok, so the problem is you doing the math wrong. Note that the MacBook Pro configuration you're talking about has a DRAM capacity of 36GB, compared to 48+ GB for the ones with all the cores enabled and the full memory bandwidth. That 32-core config isn't running the DRAM slower, it's running with a narrower bus and fewer DRAM chips: https://theapplewiki.com/wiki/MacBook_Pro_(16-inch,_M5_Max)
> There is no "package" here. Apple's processors are soldered to the logic board, as are Intel's in laptops.
Denying the difference between putting the RAM on-package vs on the motherboard doesn't make that difference stop being real.
> Apple being the first to ship LPDDR5-9600 when it was that recent doesn't imply that it needs to be soldered
Apple wasn't even close to being the first to ship LPDDR5-9600. Android phones using DRAM at that speed started shipping at the end of 2023, and moved on to 10700MT/s starting in 2024. The situation here is not anywhere close to being one of Apple paying a premium to get faster DRAM chips that other laptop manufacturers can afford. Rather, for most of the past several years, laptop manufacturers (especially on the x86 side) have been unable to buy DRAM chips with a rating slow enough to match what their processors are capable of running at. It's become quite common to see on a Thinkpad spec sheet that eg. the DRAM parts are rated for 7467MT/s but will only operate at 6400MT/s due to processor limitations, then the next year see that the DRAM parts are rated for 8533MT/s but run at 7467MT/s, and so on. LPDDR speed increases have been driven primarily by flagship smartphones, and even the leftover slower-binned parts are faster than what most laptops can handle.
> LPCAMM and similar solutions exist, but have never been demonstrated running at speeds that match what the leading soldered memory systems are using;
Does it need to be leading, though? Being median is just fine for what high-RAM systems are intended to be used for.
Don't I/you wish. The mechanical junction adds no delay, only manufacturing expense, and the delay of purchasing new systems to keep up with OS bloat.
Actually the opposite is true. Socketed RAM can be made to overclock and adjust timings, while soldered ram, no. Two Lenovo's one soldered ( Carbon X1 ), one T590, one slot: Crucial 16GB, 260-pin SODIMM, DDR4 PC4-19200. Exact same processor, the X1 is DDR3 soldered on 532.0 MHz PC3-1066. The T590, has DDR4, PC4-19200, 1200Mhz.
Both have a Core i7 8665U... and the T590 is much faster, with socketed ram.
I think you'll find that in the current day, high speed LP(?)DDR5 requires a better signal path than what the SODIMM can provide. Which is why laptop makers initially moved to soldered RAM before moving to CAMM (probably only for the high end ones).
A note that this was rather common on the days before PC clones took off.
The vertical integration many associate with Apple, was the common approach to most 8 and 16 bit home computers.
Naturally after all these years, many PC vendors want their margins back, and thus the phenomenon of everyone going back to vertical integration, especially in form factors that are ideal for such, like laptops, tablets and phones.
So the option boils down to classical desktops, or being picky on which laptops to buy.
Maybe I won't care about upgradeability right now. The architecture is clearly in flux, the roles of traditional "CPU" and "GPU" are rapidly evolving. Maybe in 5 years, or even 3 years, a brand-new machine from 2026 won't be worth upgrading for a new role due to a seriously different architecture, but would only be relegated to do something "traditional".
I wish manufacturers could consider a hybrid approach. There should be no reason an architecture can't support both unified memory (effectively L4(?) cache), and cheaper, upgradeable system memory on sticks for old-school application use.
Upgradable memory and unified memory aren't entirely mutually exclusive. You can design a chip that uses DDR5 and has a decently-powerful iGPU that can use that whole memory pool. But you'll be starving that GPU of bandwidth relative to what you'd achieve with soldered LPDDR, and it's not really worth the trouble of building a large iGPU unless you're also going to feed it with the fastest memory you can reasonably put down.
If you look at eg. an Intel laptop chip, you'll see they design and build a memory PHY that can interface with either DDR5 or LPDDR5x. They don't support splitting it to have one controller operating with DDR5 and the other with LPDDR5x, for fairly obvious reasons: more complex hardware, harder for software/operating systems to manage optimally, and not a lot of benefits to drive demand and justify the expenses. The speed difference between LPDDR5x and DDR5 isn't really large enough to use LPDDR5x as an L4 cache; it would be more like two different NUMA nodes, with complications for laptop power management.
If you want somebody to build a chip with more than the usual 128-bit bus and make some of the memory controllers use LPDDR and some DDR5, then you're asking for a significant increase in chip cost due to the extra memory PHYs and pin count. That cost is only justified if almost all products using the bigger chips are going to actually take advantage of the full complement of memory controllers.
AFAIK PCIe6 just started getting implemented in hardware last year... PCIe7 Spec was just released last year too...
PCIe6 is a much larger change than 'just bump up the transfer rate', the encoding changed too (on top of the new code length, it's no longer NRZ,) so everyone needed to design and validate both the new encoding block, negotiation, etc etc.
That said, I'm guessing PCIe7 will be a 'smoother' transition from PCIE6, i.e. we might see 7.0 products in 2027. That will theoretically get you ~240GB/sec, on an x16 link, or hypothetically a little less than the hypothetical max of a current Strix Halo. (I'm guessing however, that PCIe protocol overhead will make the difference larger.)
* 15% of British power generation is already de-linked from the gas power price. This is from 10 GW of operational renewable capacity currently covered by the price-setting Contracts for Difference scheme.
* A third (36%) of power generation in Britain will be priced independently of gas by 2030, according to forecasts of generation from the Contracts for Difference scheme. Up to 36 GW of competitively-priced new wind and solar is under development by 2032 through the scheme.
* Hours where gas was below 20% of Britain’s electricity mix averaged £60/MWh in 2025, compared to £130/MWh in hours where gas made up more than 50% of the mix.
* The gas share of power generation fell to a second record monthly low in a row in Britain in April 2026. The gas share in both March (27%) and April (19%) was the lowest per month in over a decade. In March, wind supplied a monthly record 42% of all power generation.
(As more wind, solar, batteries, and transmission are deployed in the UK, fossil gas power share will continue to decline until pushed out of the generation mix)
Looking at peak wind isn't particularly helpful I'm afraid. The problem is what happens when there is no wind, like in early May (https://gridwatch.co.uk/). Having more wind capacity isn't helping. Batteries are for smoothing minutes/hours not days. Solar obviously only a few hours a day and weak at this latitude.
Modern solar work nicely in UK in May-August when wind is weakest due to long hours and cooler weather. However one needs more expensive panels that also work on a cloudy days.
Then in UK somebody calculated that a house needs 1MWh battery to last over winter using only solar panels that a typical suburb house can install. In 5-10 years that would cost 40K USD making it rather realistic to have. This ignore availability of industrial-scale wind which is the strongest in winter.
reply