webdevver

>Isn't it a bit exaggerating to say that users cannot use Snapdragon laptops except for running LLMs?

I think maybe what OP meant was that the memory occupied by the model meant you couldn't do anything alongside inferecing, e.g. have a compile job or whatever running (unless you unload the model once you've done asking it questions.)

to be honest, we could really do with RAM abundance. Imagine if 128GB ram became like 8GB ram is today - now that would normalize local LLM inferencing (or atleast, make a decent attempt.)

ofcourse youd need the bandwidth too...

i cant believe we're still putting NPUs into new designs.

silicon taken up that couldve been used for a few more compute units on the GPU, which is often faster at inference anyway and way more useful/flexible/programmable/documented.

would be really funny if this was actually someone preparing for who wants to be a millionaire

i think i might be brain damaged because i actually like the eclipse ides they put out, atleast the stm32 one

it is an interesting thought: run a vanity fluid sim on your website to filter out everyone without discrete GPUs...

> https://wttr.in/

didnt load for me

i have an even stupider question, which is what if we scrapped ip addresses and just used ethernet addresses to route everything? just make the entire internet network be one big switch.

i think the guy who created tailscale wrote about something like this...

theres a fun "K-shaped" optionality with LLMs: on one hand, its possible to deal with otherwise large API surfaces.

but on the other hand, you can 'go oldschool' but with the hot new tools: install ubuntu, launch claude with yolo mode, and just tell it what you want as if it were a sysadmin from the early 2000s/late 90s.

both roads very reasonable, but that the old way of doing things is new again is interesting.

honestly, with LLMs, everything is fun again.

embedded dev with a billion toolchains, GPU development with each vendors bespoke API, ffmpeg with its billion parameters - if anything, you could say LLMs bailed us out of the impending ultra-specialization. without LLMs, we might be facing a world where 30% of the workforce is in software dev.

i am keeping my eyes peeled on vibe-coding PCB layouts and schematics. a lot of eyes in that direction already but its still early.

thats the one, thankyou for posting!

its fun seeing HN articles with huge upvotes but no comments, similar to when some super esoteric maths gets posted: everyone upvotes out of a common understanding of its genius, but indeed by virtue of its genius most of us are not sufficiently cognitively gifted to provide any meaningful commentary.

the karpathy vids are very cool but having watched it, for me the takeaway was "i had better leave this for the clever guys". thankfully digital carpentry and plumbing is still in demand, for now!

>How he did manage to avoid lawsuits from Microsoft is beyond me.

MS probably chose not to shut down that effort on the basis that it was enabling the MS stack in Linux.

I wish I could dig up an internal presentation that was prepared in the 90s for Bill Gates at the time, which evaluated the threat posed by Linux to Microsoft. I think they were probably happy that Linux now had a reason to talk to Windows machines.

well yes, but from todays events it seems that everybody elses military is degraded, if anything.

after being inundated with scenes of russian donkeys, ukranian 3d-printed drones, and hezbollah toyota machine-guns, this is a stark reminder of who the real, OG 'world policeman' is (and there is only one!)

ultimately its actions that matter, and the US can punch.

every dictator of every 3rd world nation is thinking super hard today about how much they really care about resisting 'US imperiaism' - indeed, whether their soldiers are even capable of preventing them from being abducted in a shock extraction by a vastly more organised and technologically superior US.

you say cannon fodder - the latest reports indicate not a single casualty in todays US offensive.

infact US combat operations are very well known for minimizing troops casulaties, and when it happens an entire movie gets made about it (black hawk down)

china has absolutely zero track record in might-makes-right style of politics.

and not to mention how many chinese students are currently betting their future on living in the US, and how many billions of dollars of chinese assets are invested in the US. if anything, chinese expats in the US would actively lobby the CCP to avoid a military confrontation with the US, whatever it takes.

nah thats cope. idk what putins issue is, i think hes just old, or maybe its just the best they can do. obviously ukraine is not venezuela but frankly 'slow and steady' is a losing strategy most of the time.

something ive read somewhere, is that the whole world is 'declining' in some sense - everywhere the median age is rising, the population is getting tired, but the US is getting tired 'less quickly' than other countries, and in that regard they are pulling ahead relatively speaking.

so youve got to get your wars in 'now' so to speak, before you don't have any young men left. in that sense, a fast and fierce offensive makes even more sense.

no way - this notion of "exchanges" isnt real, certainly i will never believe it.

why would the US 'play nice', when it can have its cake and eat it too? venezuela show of force, followed by imminent regime change in iran, ukraine will get more weapons, and china will continue getting surrounded via japan.

i think the people hoping to see some kind of change to the status-quo will be disappointed! pax americana alive and well.

video format world is one where you nope out pretty quick once you realize how many moving pieces there are.

ffmpeg seems ridiculously complicated, but infact its amazing the amount of work that happens under the hood when you do

    ffmpeg -i input.mp4 output.webm

and tbh theyve made the interface about as smooth as can be given the scope of the problem.

Down with large line counts!

LLMs have done two things for us: made lines of code a commodity, but arguably even more important: made understanding an arbitrary amount of lines of code a commodity too.

with this in hand, large codebases that were designed in the primordial era of requiring human cognition to grok are emabarassingly antiquated.

"umm what does this mean?" -> right click -> explain with AI -> "got it".

I literally reddit-soyjacked when i saw "tackle LLVM removal". I would argue that LLVM was/is the most well thought out and well-architected piece of software that exists today... but in the age of AI... who could possibly care now? when I say "well thought out" and "well-architected", that is in the context of human eyeballs. But who needs human eyeballs when we have much more powerful computer eyeballs? who grok code way better and way faster than we do.

now, the massive line count of projects like LLVM, Linux, and Chrome become much harder to defend in the context of code re-use. Why drag in millions of code when you can get an LLM to write just the bits you care about? LLVM has a ton of optimisations that are all well and good, but how many of them actually move the needle on HPC platforms that do an awful lot of instruction re-ordering and re-scheduling anyway?

Could we get away with 5% of the codebase to do "just the bits we care about"? in the past the answer might be, "well yes but its still a lot of (manual) coding", but now - why not? Let an LLM loose on the design doc and fill in the gaps!

"no no officer this is an orange pi! an orange pi!"

i am actually eagerly waiting for someone to show the real-deal: actually everything in a github repo, including 'artfiacts', or atleast those artifacts which can't be reconstructed from the repo itself.

maybe they could be encrypted, and you could say "well its everything but the encryption key, which is owned in physical form by the CEO."

theres a lot of power i think to have everything in one place. maybe github could add the notion of private folders? but now thats ACLs... probably pushing the tool way too far.

i am quietly waiting for the "bitter lesson" to hit compilers: a large language model that speaks in LLVM IR tokens that takes unoptimized IR from the frontend, and spits out an optimized version that works better than any "classical" compiler.

the only thing that might stand in the way is a dependence on reproducibility, but it seems like a weak argument: We already have a long history of people trying to push build reproducibility, and for better or worse they never got traction.

same story with LTO and PGO: I can't think of anyone other than browser and compiler people who are using either (and even they took a long time before they started using them). judged to be more effort than its worth i guess.

79 years of age!? umm... i dont understand - is this supposed to be a warning to others, or an invitation?

where was the FBI for the last 40 years? or did he really just go postal post-covid?

people say apple is often the brave company that does something unpopular that is then followed by everyone.

i think, genuinely, the brave thing for apple to do here is to cut the foldable. they could even use this leaker as an excuse to make a big splash in the news (although, maybe that will send a bad signal, that leakers have the power to end products - so maybe not.)

cut this project. nobody wants this, its a dumb gimmick aped by handset guys who are trying to differentiate their product with no real ideas or use cases.

maybe it has merit as a research project and an existence proof. but even so - where would you stick a flexible display? maybe inside the vr goggles? but who wants vr anyway? another tiny niche product with no mass market appeal.

i already left another comment on another apple thread - apple should go for the server market. offer self-hosting, but vertically integrated. do for servers what ubiquiti did for routers/networking. i keep seeing people buying mac studio clusters and hooking them up in ad-hoc thunderbolt networks. yes ofcourse social media is not 'the world', but it is a signal. I see mac minis being rack-mounted in the most janky ass looking rigs imaginable. there is a desire for this. people want servers "done right" with no corners cut, which is what apple is all about.

i get the notion that "human things should not pretend to be human", but there is funny circularity (?) here in that the "thing" that is accused of being "human" is not in fact human, and its alleged attempt to mimick human-ness is infact a projection of the human in question - which actually vindicates the non-human thing in using human language, since after all, if the squishy humans will accuse you of pretending to be human anyway (despite clearly not being one), you might aswell do so!

the article is... whatever.

BUT... on the subject of apple:

i wouldnt be surprised if apple is going to bring back server offerings. the industry is swinging towards datacenter again and apple already has historic precedent there (Xserve)

plus consumers are getting made super poor now (it is what it is), and smartphone features are getting rapidly commodified while simultaneously failing to innovate: VR, foldables, etc, these are all attempts to try and get away from getting eaten from below, and all have failed. you can be the shiniest smoothest small black square on the market, but its becoming imperceptible (another example, retina displays: every display is a retina display now.)

well, i suppose this links back to the subject - perhaps the guys at the top can see that the tech road is no longer heading in a consumer-adjacent kind of place. the serious investment will be into infrastructure and large scale stuff, not so much for the customer directly.

i would like to push back a bit on this and say, it is worth working hard, but i would argue a lot of programmers get hindered by the illusion that programming is important, or even delivering results. unfortunately in software this is exacerbated by the very real world-wide impact that programming actually does have, but: it will always be subjugated by the most important job...

the most important job, that has ever existed, and that will ever exist, is politics. moving up the career ladder you have to start thinking in terms of people, or maybe even in terms of mammals and mammalian group dynamics, cos thats who youre "programming" now, not computers. and most programmers aren't cut out for that, just as most regular people aren't cut out for programming. its hard to say why, but thats the on-the-ground data i see again and again.

i also would like to push back on the "personal projects" mindset - the sentiment often being "just live in your own little world" (not saying it is here, but this is what it often implies.) if youre going to admit defeat and retreat, be honest about what youre doing. dont dress it up as a 'win'. ceding financial/social/political agency is never a victory, but sometimes a neccesity. quitting a mag7 like google is objectively a step down whichever way you slice it. you can count on two hands the number of companies that have the level of resouces that google does - it might be worth to swallow ones pride and slog it out.

yeah thats kinda harsh, phoronix is a good oss news aggregator at the very least, and the PTS is a huge boon for "whats the best bang for buck llvm build box" type of question (which is very useful!)

the pc gaming market is a hobbyist niche compared to the ongoing infrastructure projects.

i predict that the "pc" is going to be slowly but surely eaten bottom-up by increasingly powerful SoCs.

it is surprising (or not?) that there is such a vast gulf in terms of automated tooling between the semiconductor world and pcb routing world.

i guess maybe there are less degrees of freedom and more 'regularity' in the semiconductor space? sort of like a fish swimming in an amorphous ocean vs. having to navigate uneven terrain with legs and feet. the fish in some sense is operating in a much more 'elegant' space, and that is reflected in the (beautiful?) simplicity of fish vs. all the weird 'nonlinear' appendages sticking out of terrestrial animals - the guys who walk are facing a more complicated problem space.

i guess with pcbs you have 'weird' or annoying constraints like package dimensions, via size, hole size, trace thickness, limited layer count, etc.

reminds me of this remark made by Carmack on hidden surface removal

https://www.youtube.com/watch?v=P6UKhR0T6cs&t=2315s

> "research from the 70s especially, there was tons of work going on on hidden surface removal, these clever different algorithmic ways - today we just kill it with a depth buffer. We just throw megabytes and megabytes of memory and the problem gets solved much much easier."

ofcourse "megabytes" of memory was unthinkiable in the 70s. but for us, its unthinkable to have real-time frame inferencing. I cant help but draw the parallels between our current-day "clever algorithmic ways" of drawing pixels to the screen.

I definitely agree with the take that in the grand scheme of things, all this pixel rasterizing business will be a transient moment that will be washed away with a much simpler petaflop/exaflop local TPU that runs at 60W under load, and it simply 'dreams' frames and textures for you.