Especially with anything resembling a usable amount of RAM. Mac Minis and Studios >=64GB are basically permanently sold out everywhere, because everyone, including commercial entities with deeper pockets than most of us plebs, has the exact same idea at the exact same time.
Somewhat with you on this. I got slightly excited for a brief moment, but then the site starts to scream "an LLM threw this together super quickly" which doesn't spark joy at all.
I then started digging into the code examples and quickly determined that nothing about this project is for me, even as a fan of Rust and some of its influences it has on recent languages. That web routing example is absolutely gross to my eye, for example.
Different strokes for different folks - my own thoughts on language design (I'm hacking on one in private over the past several years, maybe one day it'll be shareable) would probably make some folks have a similar reaction, despite taking a wildly different approach than here. But it does suck to see Yet Another Vibe Looking Site hosting a language that feels like Yet Another Flavor Of Similar Stuff. Really looking forward to a language that wildly shakes things up in a usable way, and has a lot of care put into the DX... this one did not check that box for me.
Using even double the total tokens and taking, what, 2-3x the time?, still seems worth it if prices are 5x+ cheaper (which OpenRouter [1] claims is the case).
On NeuralWatt for my personal projects at home (not affiliated, just a happy customer), I get so much more mileage out of GLM than I get out of Claude at work, specifically because it's priced as a hammer I can pound any nail-shaped-object with, not a delicacy I need to carefully budget-analyze to try to figure out if it's worth burning my monthly spend limits on this task.
Wildly unfriendly and standoffish, but also extremely unaccepting of dinguses blasting music without headphones on trains and buses. You win some, you lose some. I’ll embrace the silence and make my friends outside of transit and elevators, I think. (Bias disclaimer: PNW dweller for the past 9 years)
OpenCode is a quirky, buggy mess in my experience, but Pi is pretty solid and I’m glad to have recently switched over. Zerostack looks promising for certain types of users too.
4-5 bit quants would probably fit pretty well on your rig. Check HuggingFace for Qwen3.6-35B-A3B-MTP-GGUF [1]. They've also got a cool UI thing these days to help indicate which quants of a model will run on your hardware.
Full octane isn't gonna fit on much of anything south of a 128GB machine once adding KV cache.
Another sibling thread already called this out, but mentioning here: it's not "USA only", it's "US citizens only" (and I'm not entirely sure how dual-citizenship interacts with this, but I assume you can't sell to them, either, since they are by definition also foreign nationals). A private company only being able to do business with folks they can verify are solely US citizens (who themselves are also willing to submit verification of said citizenship to a private company), has a relatively small pool of potential users.
And so if this policy holds, Anthropic has functionally had Fable killed by government intervention, and in a logically consistent world, this would imply all other US-based AI labs may also never exceed existing (read: Opus) capabilities.
Coding plans are a (massive) subsidy. We can debate until the cows come home whether western frontier models' API pricing rates are fair, but the coding plans are all heavy discounts below those API rates meant to draw people in and get them hooked (and, ostensibly, to be useful for hobbyists or other lower-usage cases).
It's been discussed at length (on this site, on other sites, on like every blog ever, etc) that, eventually, those subsidies will end, much as the $5-10 Ubers/Lyfts I used to take from the far north end of Chicago into the Loop in 2016 would eventually end once those companies had a footing and didn't need to hook folks.
So - yeah, I mean, a v5 model launching in a year where Anthropic has a rather deeply established market and in a year where AI costs are rising from nearly all providers (sometimes for multiple reasons) seems like exactly the thing I'd expect them to pull the subsidy plug on after a launch teaser.
(Even the open-weight models sometimes do this: for example, OpenCode Zen/Go has a rotating door of free models at any given time that eventually leave the free tier and move into the paid tier once the launch day hype/marketing dies down)
If I pay you for a service, what implicit right should you have to then continue to profit in perpetuity by storing the data I paid you to process?
If LLMs were free your Gmail analogy might hold up. They aren’t, and so it doesn’t.
AI development can continue with the data folks opt into, or with the data AI companies incessantly scrape with reckless disregard for polite system loads. AI development does not require retaining all user inputs forever.
Windows support is huge. One of the barriers to me considering QBE for a project in recent memory was that it had no story for proprietary OSes (Windows, MacOS), and whether I like it or not, those make up the overwhelming majority of desktop-like market share. (this is the same reason I find Hare, a language that builds with QBE, interesting but not practical for my own uses - targeting only Linux and the BSDs is a non-starter, even if I personally am a Linux-only guy)
They're comparing to Haiku, not Opus. Haiku is currently at 4.5.
Even if it were Opus, comparing to a version number makes for an interesting snapshot of time comparison: if you knew how a model performed at whatever time in was in vogue, you can say "well, it looks like Model X is about 6 months/1 year/etc. behind the frontier SOTA" - which is exactly the discussion that happens in the open-weight/local LLM space. (interesting, MAI-Code-1-Flash does not appear to be such an open-weight model, following the western trend of locking models up)
My understanding from a friend who has one of these machines is that while 120Hz works, VRR does not and so the panel can’t clock down to save battery life when you’re just idling staring at a terminal.
(Not that I know all that well how good Linux machines are at clocking down anyway - my XPS and desktop both have VRR panels, but for all I know Niri runs them at full bore at all times - haven’t tried to measure, wouldn’t even know where to start)
Depending on the company, you sometimes get routed to the angry people recovery section when you do this. And so then, the Comcast agent on the other end is in stern counselor mode ready to de-escalate what seems to the robot to be a fuming, angry customer and gets completely thrown off when you’re super chill with them (or at least, you hopefully are chill with them!)
I would not. I know a handful of folks who can kinda sorta make their way through hello world in assembly with the docs open, and a handful more who could maybe implement some of the simplest coreutils like cat, maybe. But most devs I know have never seriously written a line of amd64 or aarch64 assembly. It’s just not commonly practical knowledge- even if it is very cool knowledge and helps one understand why things work (or don’t work) under the hood.
Even knowledge of how to drop to C is fairly rare in much of development, and you know what? That’s okay. We all specialize in our own areas of this beast of a field.
I only wish I could find these "a little bit empty", "80-85%" flights. Any flight I've been on in the past 4 years has been filled to the brim and gate-checking anyone who didn't pony up for Premium Economy or better - or even outright oversold and handing out "please take the next flight?" vouchers.
(Seattle is my home airport, so maybe this has something to do with it - but come on, SEA is no ORD or ATL...)
GP is clearly providing examples of categories of tasks. Sure, not all languages do “async fn foo()”, but almost all problem domains involve some sort of making sure the right things happen at the right times, which is in a similar ballpark.
Holier than thou “yeah well I work on stuff that doesn’t use databases, checkmate!” doesn’t really land - data still gets moved around somehow, and often over a network!
Oftentimes comically lower. I remember in Chicago the interstates having posted speed limits of 45mph... the average flow of traffic outside of rush hour was easily north of 70mph.
Looking even at normal arterial streets, many streets in Seattle are marked 25, but you'd be hard-pressed to find even a cop going under 30 most of the time.
I truly don't understand US road design. The construction of the road and the posted speed limit almost never are even gently correlated other than on a few select residential side streets in a few select cities who have rebuilt streets based on safety studies.
To each their own. The OS is easily one of the most frustrating I’ve ever been required to use. It does some things very well, but many things absolutely infuriatingly.
Now, yes, almost everything about Apple’s hardware UX is a light year ahead of most competitors. That’s been true for ages.
Washington State