Comment by barrkel
9 hours ago
I don't buy the central thesis of the article. We won't be in a supply crunch forever.
However, I do believe that we're at an inflection point where DC hardware is diverging rapidly from consumer compute.
Most consumers are using laptops and laptops are not keeping pace with where the frontier is in a singular compute node. Laptops are increasingly just clients for someone else's compute that you rent, or buy a time slice with your eyeballs, much like smartphones pretty much always have been.
I personally dropped $20k on a high end desktop - 768G of RAM, 96 cores, 96 GB Blackwell GPU - last October, before RAM prices spiked, based on the logic that hardware had moved on but local compute was basically stagnant, and if I wanted to own my computing hardware, I'd better buy something now that will last a while.
This way, my laptop is just a disposable client for my real workstation, a Tailscale connection away, and I'm free to do whatever I like with it.
I could sell the RAM alone now for the price I paid for it.
We won't be in a supply crunch forever. We'll have a demand crunch. The demand of powerful consumer hardware will shrink so much that producing them will lose the economics of scale. It 've always been bound to happen, just delayed by the trend of pursuing realistic graphics for games.
People who are willing to drop $20k on a computer might not be affected much tho.
> People who are willing to drop $20k on a computer might not be affected much tho.
They probably won't, but those willing to drop $3-10k will be if the consumer and data-center computing diverge at the architectural level. It's the classical hollowing out the middle - most of the offerings end up in a race-to-the-bottom chasing volume of price-sensitive customers, the quality options lose economies of scale and disappear, and the high-end becomes increasingly bespoke/pricey, or splits off into a distinct market with an entirely different type of customers (here: DC vs. individuals).
My bet is that phone hardware will be used more and more in mini PCs and laptops keeping the cost down and volume up. We see it with Apple and many Chinese mini PC makers I looked at.
The original Raspberry Pi was built around an overstock phone chip. Modern alternatives built around Rockchip and similar high-end phone chips venture into the territory of lower-end laptops. Aliexpress is full of entry-level laptops based on ARM phone chips (apparently running Android).
This will likely extend further and further, more into the "normie" territory. MS Windows is, of course, the thing that keeps many people pinned to the x64 realm, but, as Chromebooks and the Steam Deck show us, Windows is not always a hard requirement to reach a large enough market segment.
1 reply →
If this ends up being true, desktop Linux adoption might make inroads. Windows apps run like crap on ARM and no one is bothering to make ARM builds of their software.
1 reply →
All we need is for HDMI to be unlocked so it works on phones, or maybe VGA adapters that work on phones. And a way to "sideload" our own apps. Hackers please make this happen.
Unified hardware helps some and hurts some. See: same gpus for gaming and for AI.
Apple just launched a $600 amazing laptop and the top models have massive performance. What are we talking about here?
I don't think personal computers will go away, but I think the era of "put it together yourself" commodity PC parts is likely coming to an end. I think we're going to see manufacturers back out of that space as demand decreases. Part selection will become more sparse. That will drive further contraction as the market dries up. Buying boxed motherboards, CPUs, video cards, etc, will still exist, but the prices will never recover back to the "golden age".
The large PC builders (Dell, HP, Lenovo) will continue down the road of cost reduction and proprietary parts. For the vast majority of people pre-packaged machines from the "big 3" are good enough. (Obviously, Apple will continue to Apple, too.)
I think bespoke commodity PCs will go the route, pricing wise, of machines like the Raptor Talos machines.
Edit: For a lot of people the fully customized bespoke PC experience is preferred. I used to be that person.
I also get why that doesn't seem like a big deal. I've been a "Dell laptop as a daily driver" user for >20 years now. My two home servers are just Dell server machines, too. I got tired of screwing around with hardware and the specs Dell provided were close enough to what I wanted.
but I don't want a $600 amazing laptop, i want a powerful desktop x86 machine with loads of ram and disk space. As cheap as it was a couple of years ago.
3 replies →
8GB isn't an "amazing" laptop, it's a budget laptop. It's also thermally constrained quite a bit, so not even as "amazing" as it could be.
1 reply →
Reading some of the doomer comments in this thread feels like taking a glimpse into a different world.
We're out here with amazing performance in $600 laptops that last all day on battery and half of this comment section is acting like personal computing is over.
2 replies →
If they choke the consumer PC long enough the segment will die
The problem is that there is a very large incentive for three large companies to corner the market on computing components, forcing consumers to rent access instead of owning.
> We'll have a demand crunch
This is what I'm afraid of. As more stuff moves to the cloud helped in part by the current prices of HW, the demand for consumer hardware will drop. This will keep turning the vicious cycle of rising consumer HW prices and more moves to the cloud.
I can already see Nvidia rubbing their hands together in expectation of the massive influx of customers to their cloud gaming platform. If a GPU is so expensive, you move to a rental model and the subsequent drop in demand will make GPUs even more expensive. They're far from the only ones with dollar signs in their eyes, between the money and total control over customers this future could bring.
Being entirely reliant on someone else's software and hardware is a bleak thought for a person used to some degree of independence and self sufficiency in the tech world.
> I can already see Nvidia rubbing their hands together in expectation of the massive influx of customers to their cloud gaming platform.
Roblox is not popular because of its graphics. Younger gamers care more about having fun than having an immersive experience.
4 replies →
> We won't be in a supply crunch forever.
This what always happens in capitalism. Scarcity is almost always followed by glut
I don’t believe we are seeing the investments necessary that would indicate this will happen.
Memory makers, for example, have sold out their inventory for several years, but instead of investing to manufacture more, they’re shutting down their consumer divisions. They’re just transferring their consumer supply to their B2B (read AI) supply instead.
Thats likely because they don’t expect this demand to last past a few years.
4 replies →
The thing is, other than AI stuff, where does a non powerful computer limit you?
My phone has 16gigs of ram and a terabyte of storage, laptops today are ridiculous compared to anything I studied with.
I'm not arguing mind you, just trying to understand the usecases people are thinking of here.
> other than AI stuff, where does a non powerful computer limit you?
Running Electron apps and browsing React-based websites, of course.
For real. Once I've opened Spotify, Slack, Teams, and a browser about 10GB of RAM is in use. I barely have any RAM left over for actual work.
22 replies →
Companies love externalizing the costs of making efficient software onto consumers, who need to purchase more powerful computing hardware.
16 replies →
It seems like as hardware gets cheaper, software gets more bloated to compensate. Or maybe it’s vice versa.
I wonder if there’s a computer science law about this. This could be my chance!
5 replies →
That's actually a good point, haha. The worst-case scenario of computers being thin clients for other people's servers dissolves when you realize that chromium/electron IS, nominally, a thin client for HTTP servers, and it'll gladly eat up as much memory as you throw at it. In the long term, modulo the current RAM shortage, it turns out it's cheaper to ship beefy hardware than it is to ship lean software.
This is the way
The big one for me is ballooning dependency trees in popular npm/cargo frameworks. I had to trade a perfectly good i9-based MacBook Pro up to an M2, just to get compile times under control at work.
The constant increases in website and electron app weight don't feel great either.
3D CAD/CAM is still CPU (and to a lesser extent memory) bound --- I do joinery, and my last attempt at a test joint for a project I'm still working up to was a 1" x 2" x 1" area (two 1" x 1" x 1" halves which mated) which took an entry-level CAM program some 18--20 minutes to calculate and made a ~140MB file including G-code toolpaths.... (really should have tracked memory usage....)
That sounds like pretty degenerate behavior. I typically have CAM toolpaths generate in seconds using Fusion or PrusaSlicer.
Is that by convention or is there a good reason that it’s so CPU bound? I don’t have experience with CAD, so I’m not sure if it’s due to entrenched solutions or something else.
2 replies →
> My phone has 16gigs of ram and a terabyte of storage, laptops today are ridiculous compared to anything I studied with.
Most affordable laptops have exactly that, 16gigs of ram and a terabyte of storage. Think about THAT!
I've never have a personal computer that came even close to powerful enough to do what i want. Compiles that take 15 minutes, is really annoying for instance.
>My phone has 16gigs of ram and a terabyte of storage
That's "non powerful" to you?
The opposite. I meant that if this is what consumer grade looks like nowadays, even with a fraction of current flagships we seem well covered - this was less than 800 bucks.
I’d love it if a clean build and test on the biggest project I work in would finish instantly instead of taking an hour.
[dead]
> "I personally dropped $20k on a high end desktop"
This absolutely boggles my mind. Do you mind if I ask what type of computing you do in order to justify this purchase to yourself?
Any and all. It's not particularly justifiable. It's more like, I'm a software engineer, and this is my home workshop. I run dozens of services, experiment with a bunch of different LLMs, tune my Postgres instance for good performance on large datasets, run ML data prep pipelines. All sorts really.
I'm also into motorcycles. Before I owned a house with a garage, I had to continuously pack my tools up and unpack them the next day. A bigger project meant schlepping parts in and out of the house. I had to keep track of the weather to work on my bikes.
Then, when I got a house, I made sure to get one with a garage and power. It transformed my experience. I was able to leave projects in situ until I had time. I had a place to put all my tools.
The workstation is a lot like that. The alternative would be renting. But then I'd spend a lot of my time schlepping data back and forth, investing in setting things up and tearing them down.
YMMV. I wouldn't dream of trying to universalize my experience.
I'm thinking the same. My total computing purchases in the last 25 years, including desktops, laptops, monitors, phones, and tablets is way under 20k.
I would bet it continues to be more affordable to buy reasonable specs with current consumer hardware, rather than buying a top system once.
I haven't purchased a new computer in, at least, 10 years. I take pride (i.e., I have a sickness) in purchasing used laptops off eBay, beefing them up, and loading Debian on them. My two main computers are a Dell E5440 and a Lenovo ThinkPad T420. I, too, am a software developer, but [apparently] not as much of a rock star software developer at this gentleman. :-D
> I personally dropped $20k on a high end desktop - 768G of RAM, 96 cores, 96 GB Blackwell GPU - last October, before RAM prices spiked […]
768GB of RAM is insane…
Meanwhile, I’ve been going back and forth for over a year about spending $10k on a MacBook Pro with 128GB. I can’t shake the feeling I’d never actually use that much, and that, long term, cloud compute is going to matter more than sinking money into a single, non-upgradable machine anyway.
Your battery is going to suffer because of the extra ram as well.
I don't know your workloads, but for me personally 64 GB is the ceiling buffer on RAM - I can run entire k8s cluster locally with that and the M5 Pro with top cores is same CPU as M5 Max. I don't need the GPU - the local AI story and OSS models are just a toy for my use-cases and I'm always going to shell out for the API/frontier capabilities. I'm even thinking of 48 config because they already have those on 8% discounts/shipped by Amazon and I never hit that even on my workstation with 64 GB.
> Your battery is going to suffer because of the extra ram as well.
No, it won't. The power drain of merely refreshing DRAM is negligible, it's no higher than the drain you'd see in S3 standby over the same time period.
2 replies →
I thought my Z620 with 128GB of RAM was excessive! Actually, HP says they support up to 192GB of RAM, but for whatever reason the machine won't POST with more than 128GB (4Rx4) in it. Flawed motherboard?
Look at the way age gating is going in a global coordinated push. Can control of compute be far behind?
It wasn't my primary motivator but it hasn't made me regret my decision.
I hummed and hawed on it for a good few months myself.
Just look at ITAR and the various attempts at legislating 3D printing and CNC machining of firearms parts to see one justification point of that.
> Can control of compute be far behind?
How is this going to work? You need uncontrolled compute for developing software. Any country locking up that ability too much will lose to those who don't.
3 replies →
> 768GB of RAM is insane.
Before this price spike, it used to be you could get a second-hand rack server with 1TB of DDR4 for about $1000-2000. People were massively underestimating the performance of reasonably priced server hardware.
You can still get that, of course, but it costs a lot more. The recycling company I know is now taking the RAM out of every server and selling it separately.
Apple hardware is incredibly overpriced.
> spending $10k on a MacBook Pro with 128GB.
As someone who just bought a completely maxed out 14" Macbook Pro with an M5 Max and 128GB of RAM and 8TB SSD, it was not $10k, it was only a bit over $7k. Where is this extra $3k going?
Tangential, I bought a nearly identically-spec'd (didn't spring for the 8 TB SSD - in retrospect, had I kept it, I would've been OK with the 4 TB) model, and returned it yesterday due to thermal throttling. I have an M4 Pro w/ 48 GB RAM, and since the M5 Max was touted as being quite a bit faster for various local LLM usages, I decided I'd try it.
Turns out the heatsink in the 14" isn't nearly enough to handle the Max with all cores pegged. I'd get about 30 seconds of full power before frequency would drop like a rock.
1 reply →
It could be a different country?
With the way legislation is going these days, self hosting is becoming ever more important. RAM for zfs + containers on k3s doesn't end up being that crazy if you assuming you need to do everything on your own. (at home I've got 1 1tb ram machine, 1 512gb, 3x 128gb all in a k3s cluster with some various gpus about about a half pb of storage before ~ last sept this wasn't _that_ expensive to do)
My home server has 512GB RAM, 48 cores, my 4 desktops are 16 cores 128GB, 4060GPU each. Server is second hand and I paid around $2500 for it. Just below $3000 price for desktops when I built them. All prices are in Canadian Pesos
Canadian Pesos?
3 replies →
> Most consumers are using laptops and laptops are not keeping pace with where the frontier is in a singular compute node.
How can you say this when Apple is releasing extremely fast M5 MacBook Pros? Or the $600 MacBook Neo that has incredible performance for that price point?
Even x86 is getting some interesting options. The Strix Halo platform has become popular with LLM users that the parts are being sold in high numbers for little desktop systems.
They're ultimately laptops, you won't be able to squeeze out the same amount of performance from a laptop compared to a desktop, regardless of the hardware.
If you haven't tried out a desktop CPU in a while, I highly recommend you giving it a try if you're used to only using laptops, even when in the same class the difference is obvious.
I have a recent MacBook Pro and a high end Zen 5 desktop.
For CPU-bound tasks like compiling they’re not that different. For GPU tasks my desktop wins by far but it also consumes many times more power to run the giant GPU.
If you think laptops are behind consumer desktops for normal tasks like compiling code you probably haven’t used a recent MacBook Pro.
1 reply →
They're fast, but they'll never even remotely reach what a mid-range desktop PC with dedicated graphics burning 500W is able to do.
A 300W GPU released in 2025 is about 10x M5 perf. The difference is going to be smaller for CPU perf, but also not close.
> The difference is going to be smaller for CPU perf, but also not close.
This is not true. The recent MacBook Pros are every bit as fast as my Zen 5 desktop for most tasks like compiling.
For GPU there is a difference because both are constrained by thermal and power requirements where the desktop has a big advantage.
For CPU compute, the laptop can actually be faster for single threaded work and comparable for multi threaded work.
Anyone claiming laptop CPUs can’t keep up with desktop CPUs hasn’t been paying attention. The latest laptops are amazing.
4 replies →
We are on borrowed time, most of the world is running on oil and this resource is not unlimited at all. A lot of countries have gone past their production peak, meaning it's only downhill from here. Everything is gonna be more costly, more expensive, our lavish "democracies" lifestyles are only possible because we have (had) this amazing freely available resource, but without it it's gonna change. Even at a geopolitical scale you can see this pretty obviously, countries that talked about free market, free exchange are now starting to close the doors and play individually. Anyways, my point is, we are in for decades, if not a century of slow decline.
Doubt it. Renewables are expanding much faster than oil output is decreasing. Wind and solar will enable energy to remain cheap everywhere that builds it.
Energy production is only part of the bill, though. The oil shortage is having an effect on a mind-boggling variety of consumer goods where crude oil is used in manufacturing. For many products we don't have good alternatives. A lot of oil is needed to build an electric car.
Malthusians has been sounding the alarm for longer than Protestant revivalists have been claiming the end of world is next month at lunchtime. If there is a predication market for such things, betting on any Malthusian is patently foolish.
(Of course, I don't disagree with the notion that consumerism produces an extraordinary amount of worthless trash, but that's a different matter. The main problem with consumerism is consumerism itself as a spiritual disease; the material devastation is a consequence of that.)
1 reply →
There will be very dramatic growing pains with this switch, especially for A: nations manufacturing renewables but still running that manufacturing on oil and B: nations that face political and economic barriers for renewables.
Also C: nations that are both A and B, needlessly causing oil volatility with unplanned military dickheadedness.
Renewables provide electricity only, but planes, boats, trucks, basically all the supply chain, works with oil only for the moment. The ease of use of oil has not been replaced yet. Do you realize how easy it is to handle oil ? You can just put it in a barrel and ship it anywhere in that barrel. No need for wires or complex batteries like for electricity, nor complex pipelines like for gas.
And even if we figured out how to electrify everything (which we didn't as I just said), we would still run into resources shortages for batteries, wires (copper etc.), nuclear fuel (uranium)...
5 replies →
Hilarious how comments like this consistently get downvoted, theres a lot of special interest lurkers on this forum
"I personally dropped $20k on a high end desktop . . . "
This is where I think current hackers should be headed. I grew up with lots of family who were backyard mechanics, wrenching on cars and motorcycles. Their investment in tools made my occasional PC purchase look extremely affordable. Based on what I read, senior mechanics often have five-figure US dollar investments in tools. Of course, I guess high quality torque wrenches probably outlast current GPU chips? I'd hate to be stuck making a $10K investment every 24 months on a new GPU . . .
I have been renting GPU resources and running open weight models, but recently my preferred provider simply doesn't have hardware available. I'm now kicking myself a little for not simply making a big purchase last fall when prices were better.
> Most consumers are using laptops and laptops are not keeping pace with where the frontier is in a singular compute node. Laptops are increasingly just clients for someone else's compute that you rent, or buy a time slice with your eyeballs, much like smartphones pretty much always have been.
It really feels like we're slowly marching back to the era of mainframe computers and dumb terminals. Maybe the democratization of hardware was a temporary aberration.
It seems like you largely agree with the article - people shall own nothing and be happy. Perhaps the artificially induced supply crunch could go on indefinitely.
Also, I wonder how many of us, even here on HN, have the ability to spend that amount of money on computer for personal use. Frankly I wouldn't even know what to do with all the RAM - should I just ramdisk every program I use and every digital thing I made in the last five years?
Anyhow, I suppose for the folks who can't afford hardware (perhaps by design), one ought to own nothing and be happy.
People spend a lot more than that on a car they use less, especially if they're in tech.
The RAM choice was because I have never regretted buying more RAM - it's practically always a better trade than a slightly faster CPU - and 96GB DIMMs were at a sweet spot compared to 128GB DIMMs.
That, and the ability to have big LLMs in memory, for some local inference, even if it's slow mixed CPU/GPU inference, or paged on demand. And if not for big LLMs, then to keep models cached for quick swapping.
I bought a 4 year old car for significantly less than that. And I can get a computer that can do 99% of what your monster can do for like 10% of the price. And if I want LLM inference I can get that for like $20 a month or whatever.
I don't mean to judge, it's your money but to me it seems like an enormous waste. Just like spending $100k on a car when you can get one for $15k that does pretty much exactly the same job.
2 replies →
Solarpunk + https://permacomputing.net/
That’s for everyone
I've always went way over on RAM, for the most part. 32, 64, then 128GB of memory.
Never really used it all, usually only about 40%, but it's one of those better to have than not need, and better than selling and re-buying a larger memory machine (when it's something you can't upgrade, like a Mac or certain other laptops)
I believe superficially speaking you could be right. But I think it was realised that causing the scarcity of products and commodities is a power move.
We live in world where we optimised for globalization. Industry in china, oil in middle east, etc...
This approach proved to be fragile on the hands of people with money and/or power enough to tilt the scale
It's not a power move, it's a cartel and they've done this before. Gamers Nexus did a fantastic piece on how where we're at today is very similar to the DRAM price fixing and market manipulation just a couple decades ago [0]. This is the big players taking full advantage of an opportunity for profit.
[0] https://youtu.be/jVzeHTlWIDY?si=cRJ6C7jPxLIpKTyF
This will be me. Bestowing upon my descendants a collection of Mighty Beanz, a few unkillable appliances, and the best consumer computing hardware the early 2020s could buy.
And I fear they will be equally confused and annoyed by disposing of all of them.
>we're at an inflection point where DC hardware is diverging rapidly from consumer compute.
I thought the trend is the opposite direction, with RTX 5x series converging with server atchitectures (Blackwell-based such as RTX 6000 Pro+). Just less VRAM and fewer tensor cores, artificially.
Where is the divergence happening? Or you don't view RTX 5x as consumer hardware?
Blackwell diverges within Blackwell itself… SM121 on the GB10 vs the RTX 5000 consumer vs the actual full fat B100 hardware all have surprisingly different abilities. The GB10 has been hamstrung by this a bit, too.
You're responding to an LLM authored article that doesn't know anything. "Let that sink in for a moment."
I think you're probably right, but I'm not so confident the supply crunch will end.
Tech feels increasingly fragile with more and more consolidation. We have a huge chunk of advanced chip manufacturing situated on a tiny island off the coast of a rising superpower that hates that island being independent. Fabs in general are so expensive that you need a huge market to justify building one. That market is there, for now. But it doesn't seem like there's much redundancy. If there's an economic shock, like, I dunno, 20% of the world's oil supply suddenly being blockaded, I worry that could tip things into a death spiral instead.
$20k?
People laugh at young men for looksmaxxing. And then there’s this. I dunno. As someone who has been playing computer games since the 70s, I clearly do not understand the culture anymore. But what forces would drive a young man to spend the price of a used car to play a derivative FPS? It seems heartbreaking. Just like the looksmaxxer.
Alas, I'm not a young man any more. And my HEDT is headless, it has no monitor with which to play FPSes.
[dead]
> Laptops are increasingly just clients for someone else's compute
Are you kidding? Apple's mobile chips are now delivering perf that AMD & intel desktop never could or did.
> Apple's mobile chips are now delivering perf that AMD & intel desktop never could or did.
Most applications don't make aggressive use of the SIMD instructions that modern x86 chips offer, thus you get this impression. :-(
Users do not care why the perf they get is what they get. What good is AVX2048 if nobody uses it?
> Laptops are increasingly just clients for someone else's compute that you rent, or buy a time slice with your eyeballs, much like smartphones pretty much always have been.
What are you talking about?
My laptops are, and always have been, primarily places where I do local computing. I write code there, I watch movies there, I listen to music there, I play games there...all with local storage, local compute, and local control (though I do also store a bunch of my movies on a personal media server, housed in my TV stand, because it can hold a lot more). My smartphone is similar.
If you think that the vast majority of the work most people do on their personal computers is moving to LLMs, or cloud gaming, then I think you are operating in a pretty serious bubble. 99.9% of all work that most people do is still best done locally: word processing, spreadsheets, email, writing code, etc. Even in the cases where the application is hosted online (like Google Docs/Sheets), the compute is still primarily local.
The closest to what you're describing that I think makes any sense is the proliferation of streaming media—but again, while they store the vast libraries of content for us, the decoding is done locally, after the content has reached our devices.
It doesn't matter if a cutting-edge AI-optimized server can perform 10, 100, or 1000 times better than my laptop at any particular task: if the speed at which my laptop performs it is faster than I, as a human, can keep up (whatever that means for the particular task), then there's no reason not to do the task locally.
> We won't be in a supply crunch forever.
I don't share the same 1:1 opinion with regards to the article, but it is absolutely clear that RAM prices have gone up enormously. Just compare them. That is fact.
It may be cheaper lateron, but ... when will that happen? Is there a guarantee? Supply crunch can also mean that fewer people can afford something because the prices are now much higher than before. Add to this the oil crisis Trump started and we are now suddenly having to pay more just because a few mafiosi benefit from this. (See Krugman's analysis of the recent stock market flow of money/stocks.)
The increase looks higher because we were at an all-time price low. RAM has been this expensive at least twice before, and it always dropped way down again after.
General predictions are in 3-5 years things will return to normal. 3 years if the current AI crunch is a short term thing, 5 years if it isn't and we have to build new RAM factories.
Local is a dead end.
Open source efforts need to give up on local AI and embrace cloud compute.
We need to stop building toy models to run on RTX and instead try to compete with the hyperscalers. We need open weights models that are big and run on H200s. Those are the class of models that will be able to compete.
When the hyperscalers reach take off, we're done for. If we can stay within ~6months, we might be able to slow them down or even break them.
If there was something 80-90% as good as Opus or Seedance or Nano Banana, more of the ecosystem would switch to open source because it offers control and sovereignty. But we don't have that right now.
If we had really competitive open weights models, universities, research teams, other labs, and other companies would be able to collaboratively contribute to the effort.
Everyone in the open source world is trying to shrink these models to fit on their 3090 instead, though, and that's such a wasted effort. It's short term thinking.
An "OpenRunPod/OpenOpenRouter" + one click deploy of models just as good as Gemini will win over LMStudio and ComfyUI trying to hack a solution on your own Nvidia gaming card.
That's such a tiny segment of the market, and the tools are all horrible to use anyway. It's like we learned nothing from "The Year of Linux on Desktop 1999". Only when we realized the data center was our friend did we frame our open source effort appropriately.
> We need open weights models that are big and run on H200s.
We have this class of models already, Kimi 2.5 and GLM-5 are proper SOTA models. Nemotron might also release a larger-sized model at some time in the future. With the new NVMe-based offload being worked on as of late you can even experiment with these models on your own hardware, but of course there's plenty of cheap third-party inference platforms for these too.
> Open source efforts need to give up on local AI and embrace cloud compute.
Oh god no, please not more slop, you're already consuming over 1 percent of human energy output, could you, like, chill a bit?
In a similar vein: seek efficiency.
I.e., /if/ I am going to consume LLM tokens, I figure that a local LLM with 10s of billions of parameters running on commodity hardware at home will still consume far more energy per token than that of a frontier model running on commercial hardware which is very strongly incentivized to be as efficient as possible. Do the math; it isn't even close. (Maybe it'd be closer in your local winter, where your compute heat could offset your heating requirements. But that gets harder to quantify.)
Maybe it's different if you have insane and modern local hardware, but at least in my situation that is not the case.
4 replies →
Y'all aren't seeing the same future I am, I guess.
- Our career is reaching the end of the line
- 99.9999% of users will be using the cloud
- if we don't have strong open source models, we're going to be locked into hyperscaler APIs for life
- piddly little home GPUs don't do squat against this
Why are you building for hobby uses?
Build for freedom of the ability to make and scale businesses. To remain competitive. To have options in the future independent of hyperscalers.
We're going to be locked out of the game soon.
Everyone should be panicking about losing the ability to participate.
Play with your RTXes all you like. They might as well be raspberry pis. They're toys.
Our future depends on our ability to run and access large scale, competitive, open weights. Not stuff you run with LM Studio or ComfyUI as a hobby.
2 replies →
Man, going to personal computing was a mistake, we should’ve stayed jacked to the mainframes /s
Entire device categories, like smartphones, are locked down. That's our future.
Here's my retort: https://news.ycombinator.com/item?id=47543367