walrus01

15 ms·

421.
by walrus01 2mo ago
Did something like the android 'osmand' app with offline openstreetmap data for .IL downloaded function?
422.
by walrus01 2mo ago
You should compare notes with the trove of weird emails that Daniel at haxx.se gets (the curl guy) because his email address is in the curl license text appended in the fine print of a bunch of products. https://daniel.haxx.se&#x
423.
by walrus01 2mo ago
probably the best experience would be deepseek v4 flash 0731 (it takes about 170GB RAM on the server side for the full thing and RAM reserved for 1M context) via opencode's $10 a month plan until you use that up, it's either Q8 or
424.
by walrus01 2mo ago
there's metadata in a GGUF (as can be seen when loading one into llama-server with verbose level 4) but it doesn't appear to show what the expected sha256 of the file is.
425.
by walrus01 2mo ago
I have just run a sha256 checksum on both copies now, the one I downloaded 4-5 days ago, and the one that's on the unsloth huggingface page released today, and will see if they're the same or different. (downloaded in the last few
426.
by walrus01 2mo ago
Yes, google "llama-server split model multiple GPU" for some concrete examples.
427.
by walrus01 2mo ago
It would be nice if unsloth published GGUFs would use a version number or something, because now I have multiple different files on local storage that otherwise have exactly the same name. "Qwen3.8-27B-UD-Q8_K_XL.gguf" for instanc
428.
by walrus01 2mo ago
Q2 quantization is basically giving a capable model a lobotomy. It will not accurately represent how smart or capable something like qwen 3.8 27B in Q8 will be.
429.
by walrus01 2mo ago
It's not that I haven't given it a try, my primary desktop machine is debian stable and xfce4 with the libraries installed to also run a bunch of gnome and kde-native applications, but that right now I prefer macos for laptop use.
430.
by walrus01 2mo ago
yeah, but if only suspend/resume worked near flawlessly... I've had MacOS crash on suspend or resume before, but I'd say it happens less than 1/10th of 1% of the time I sleep or resume my laptop.
431.
by walrus01 2mo ago
yes, though the level of headache is relatively minimal. I meant I'd stop using it if things became completely impossible to run unsigned code (or the hassle went from a couple of extra seconds of work to some convoluted process) . Add
432.
by walrus01 2mo ago
> If I can download and run arbitrary code on my Mac--even if I have to jump through scary warnings--why should I not be able to do so on my phone? I just bought a new m5 macbook air. The day that I can't treat my MacOS laptop as
433.
by walrus01 2mo ago
As an analogy, you're not going to cut your lawn by hand at the speed and slow pace that the smallest and weakest lawn mowing robot does. You're not going to cut your lawn with kitchen scissors. But the LLM will, metaphorically, b
434.
by walrus01 2mo ago
LLMs don't get as mentally exhausted by trying repeated tedious things as a human will. Same general idea as, imagine if you as a human typing into a keyboard had to manually fuzz test software in a pre-LLM era vs. how fuzz testing is
435.
by walrus01 2mo ago
One of the reasons why people just go "meh, whatever, good enough" and install wordpress, use some popular wordpress contact form. Have it give mail to a postfix daemon on same server that listens only on localhost and relays outb
436.
by walrus01 2mo ago
Only significant change I would make is have the top navbar not be transparent with body content underlaid beneath it, which makes things hard to read when the page is scrolled. And move some of the links at the very bottom of page like &#x
437.
by walrus01 2mo ago
> It's 1998: hey, I can drastically speed up your web service, let's etch it right to silicon. I distinctly remember 32-bit/33 MHz PCI accelerator cards for SSL being a real thing (for use on OpenBSD or FreeBSD), in an era
438.
by walrus01 2mo ago
The author may very well be working in Mandarin or something and asked it to create an english language explanation of the project. Then the LLM generated the text based on their request.
439.
by walrus01 2mo ago
As LLM generated template web GUI go it's relatively inoffensive, at least it's easy to navigate and doesn't waste time with sliders, animated effects, etc. It's just a navbar across the top and then content. As far as t
440.
by walrus01 2mo ago
I heard a great deal of noise from early 2002 to the present date that the US military had a high interest in small portable nuclear reactors for large bases in Iraq and Afghanistan. And particularly around the peak period of troops on the
441.
by walrus01 2mo ago
> I do not rely on any LLM of any size for general knowledge baked into the weights You have to rely on it to a certain level for agentic/coding work, presuming that's the general subject we're talking about here... For in
442.
by walrus01 2mo ago
It did OK on schlongbench v1.0 (test of a specific niche word that doesn't make it into smaller LLMs) but it sure does love to count words https://pastes.io/r8F1AY8h
443.
by walrus01 2mo ago
If this is true, it's even more impressive that some of the open weight models that are <3.5T in size, approx 33% of its size, are within a few points of it in the artificial analysis leaderboard.
444.
by walrus01 2mo ago
Without having any inside information, one possible theory: All or a vast majority of of the cerebras manufacturing capacity was going to a few companies that aren't publicly available inference providers on openrouter, for their own i
445.
by walrus01 2mo ago
Information about RAM type/size and connection topology of the RAM to be used for context cache seems to be conspicuously absent from the slick looking marketing materials.
446.
by walrus01 2mo ago
I guess we know why there's a fair bit of investment money going into small modular nuclear reactor startups now.
447.
by walrus01 2mo ago
It's mandatory liquid cooling, so it's meant to be attached to a specialized liquid cooling loop that gets the heat outside the building. This is far beyond the practical maximums of like 10 to 15kW per 44U cabinet front to rear a
448.
Running Ultrix 4.5 (2019)(astr0baby.wordpress.com)
2 points by walrus01 2mo ago|0 comments
449.
by walrus01 2mo ago
Xenix was targeted at a hardware spec that was well out of reach for the ordinary home PC consumer. The amount of ram and HDD needed to install and run it would make a computer in 1985/1986 money equivalent to a $12,000 desktop today.
450.
by walrus01 2mo ago
Buffalo Buffalo buffalo Buffalo buffalo?

More ›