kristjansson

7 ms·

1.
by kristjansson 7d ago
The user cares about the distribution of outputs. That distribution is structurally determined by the distribution of inputs (i.e. prompts), the weights of the model, and (these days) the dynamics of the harness guiding successive generati
2.
by kristjansson 12d ago
Well, time to set down the glass beads and dive into a an alpine lake.
3.
by kristjansson 13d ago
With MTP and FP4 I max out at 30ish t/s on mine. Without MTP or in regimes where the drafter performs poorly it’s about 10 t/s. FP8 is about half that
4.
by kristjansson 13d ago
They made the coding plan a bit better toward the end, but it was pretty tough to use throughout. Seems like an Amdahl’s law of inference economics? there’s so much compute relative to SRAM on the chip and shoreline bandwidth onto the chip
5.
by kristjansson 14d ago
> one TLD does not operate at the most fundamental level like the others; contrary, this is one TLD _ceasing_ its idiosyncratic mode of operation.
6.
by kristjansson 21d ago
Of course there's some trim that has to go into the grinder, but we could do a much better, more respectful job of using the animal as it was instead of homogenized lumps of 80/20 or 90/10. More beef cheek in the butcher case
7.
by kristjansson 21d ago
I don't know if you're a bit out of the loop, but ... meat is very much not at issue in this case.
8.
by kristjansson 21d ago
> grinder There's your problem. Ground products are completely dispensable.
9.
by kristjansson 23d ago
> LLMs could This is going to end up like the Law of Headlines, isn't it? "Do x, y, z Cure All That Ails You?" ... no but we got you to read the article. "LLMs _could_ x, y, z" ... but they don't because th
10.
by kristjansson 23d ago
It seems impossible for this to be a true statement. and yet. What’s next? $20m series B in ByBy, a horizontally integrated trash disposal, concrete shoe manufacture, lye bath, and pig farm operation?
11.
by kristjansson 23d ago
The operative part of schoolwork is … not the work part? Of course school work is bullshit, in the sense that the teacher already knows what 7x9 is or who fought at Waterloo or whatever. The point is not the work product, it’s that the stu
12.
by kristjansson 23d ago
“Passive Income? Bro, I need MASSIVE INCOME” as the tagline of a sports gambling operation … I guess “your desperation is our opportunity, give us your money you illiterate innumerate peasant rube fuck you” didn’t fit on the billboards?
13.
by kristjansson 27d ago
hf download maintains each downloaded version in a commit-hash-addressed snapshots dir, with symlinks to the underlying files content-addressed in a blobs dir. It’s not quite git, but it is a stable reference path.
14.
by kristjansson 28d ago
Just quantizing takes seconds-to-minutes, llama.cpp provides a nice tool[0]. Improving quality is then a matter of picking specific tensors to maintain at higher accuracy, checking on representative data, and repeating. [0]: https:/&
15.
by kristjansson 1mo ago
I'm generally happy with the tests I ask it for, some of which are PBT. It's just the insistence upon memorializing every single change with a test. Maybe encouraging / forcing PBT will dissuade it?
16.
Expert Witness to ChatGPT: "Show how 3M is 0 percent at fault"(404media.co)
75 points by kristjansson 1mo ago|45 comments
17.
by kristjansson 1mo ago
And the tests. Oh god the tests. Personal recent favorite: I asked for some changes to a Dockerfile, which it did ably, and then promptly tested by writing a pytest module that traversed up to the root, read the Dockerfile, and checked that
18.
by kristjansson 1mo ago
A big differentiator is whether one imagines an LLM in-band with most/all future software. If there is, and we’re deferring until very late parts of a program that world have been load bearing, and we’re able to programmatically ands r
19.
by kristjansson 1mo ago
the norms of issue trackers are strongly opposed to “bump”. these autoclose bots may yet change that
20.
by kristjansson 1mo ago
harnesss == thing that calls LLM API, acts on response, and maybe does that again.
21.
by kristjansson 1mo ago
> Deepseek v4 pro 0813 Which itself released yesterday? You're writing, reading, and evaluating enough software in a ~36 hour period to form, reject, and form another opinion about which model makes better _architectural_ choices?
22.
by kristjansson 1mo ago
Oh shoot it had a nice countdown timer to ~8a PT when I submitted it
23.
Qwen3.8 27B(huggingface.co)
65 points by kristjansson 1mo ago|13 comments
24.
by kristjansson 1mo ago
How they have any credibility after they told everyone they could run Deepseek R1 on their laptop (by giving the Qwen 7B distill the `deepseek` moniker)...
25.
by kristjansson 1mo ago
> atem also perhaps taking some small joy from the lexical similarity to aten[0] namespace that lies at the heart of pytorch [0]: https://github.com/pytorch/pytorch/blob/main/aten/src/README.
26.
by kristjansson 1mo ago
including Anthropic[0] and OAI[1] [0]: https://platform.claude.com/docs/en/agents-and-tools/tool-us... [1]: https://developers.openai.com/api/docs/guides/tools-programm...
27.
by kristjansson 1mo ago
As a relatively recent herdr user, I’ll say (1) congratulations and (2) the thing that was most attractive was the seeming orthogonality of the tool. It sits apart, interfaces with subordinate programs in a few clearly delineated ways, and
28.
by kristjansson 1mo ago
Also: your messages causing your LLM (harness) to run CLIs on your computer? charming, thrilling, great fun. other people’s messages causing your LLM to run CLIs on your (cloud) computer? terrifying, awful, sickening, no fun at all
29.
by kristjansson 2mo ago
Apropos of nothing: https://en.wikipedia.org/wiki/Marcus_Licinius_Crassus
30.
by kristjansson 2mo ago
> An orange frisbee is harder to find in grass then a red one same for ungulates. hence, tigers

More ›