jchw

6 ms·

31.
by jchw 1mo ago
Why do you care so much that this is a totally realistic argument? You don't personally have to like it, it's not about how it makes you feel as a fan of Apple.
32.
by jchw 1mo ago
Or an agency on their behalf, but either way, there's really no way for us to really know, due to the stealthy nature of this sort of advertising.
33.
by jchw 1mo ago
To be fair, the very reasonable take is that if there was guerilla marketing campaigns, they are most likely being handled by external agencies and not Apple directly. And, this isn't making them look particularly dumb. It is making
34.
by jchw 1mo ago
You're suggesting there is no chance that Apple is engaging in guerilla marketing, which you discredit by referring to as a "psyop" when it is in fact a well-documented and pretty mundane practice, because they sell a lot of
35.
by jchw 1mo ago
I think its mere existence and influence is already good enough. A fair number of modern open source projects either come with Nix flakes or are designed to work well in an immutable environment whether it is NixOS, SteamOS, or inside of a
36.
by jchw 1mo ago
I'm not entirely sure if it's a good idea per se, but hell, what's the point in making everything properly lazy if you couldn't. I mean me personally, I routinely use 20-30 flake inputs for large system configurations an
37.
by jchw 1mo ago
When people pointed out the consequences this would have for open source, that did not mean what they wanted was an open source exemption. It was just meant to demonstrate just how poorly conceived the law is that nobody at any point even t
38.
by jchw 1mo ago
Ah, I see - so MoE offloading is no savior. A shame but no surprise either.
39.
by jchw 1mo ago
One thing I'd like to try is MoE offloading: I have 2x32 GiB of VRAM and 128 GiB of DDR5 running at 4800 MT/s (only 2 channels though). I've seen people post difficult to believe MoE offloading results albeit a decently long
40.
by jchw 1mo ago
Honestly I suspect neither of them will be performing terribly well but with DDR4 3200 RAM I wonder if you'll be counting tokens per second or seconds per token. I mean, you do at least get a lot of memory channels at least, compared t
41.
by jchw 1mo ago
See the note at the bottom here: https://news.ycombinator.com/newsguidelines.html
42.
by jchw 1mo ago
Well, you know what they say. The first one's free.
43.
by jchw 2mo ago
Almost 10 models are passing the benchmark >95% - isn't that... substantially overly saturated? I'm also really skeptical of benchmarks that place any Haiku model very high. I've been thoroughly unimpressed with Haik
44.
by jchw 2mo ago
Cool. I was thinking of running Qwen3.8 through Codex, but maybe it's time I take a look at Pi.
45.
by jchw 2mo ago
Seems this person is well known for saying silly things based on the comments but I read it carefully and tried to give it a fair shake. > Accusations of mental instability are a favorite of fascist regimes, so if you see that you know i
46.
by jchw 2mo ago
You are misunderstanding what they said, they are saying you can use uncensored variants of models like Qwen when running locally. There are quite a lot of people working to "uncensor" open weights releases. It seems to work altho
47.
by jchw 2mo ago
I'd personally like to know more about what tools it used/wanted and the harness setup, because this sounds pretty cool. I have a dual Arc Pro B70 setup and currently get around 22 t/s which isn't great but isn't
48.
by jchw 2mo ago
Despite all of this though, GPT 5.6 Sol to me has significantly less trouble with this. It still suffers from LLMisms to some extent (I hold that this is probably due to human feedback in training just doing a bad job for prose) but I defin
49.
by jchw 2mo ago
You are confusing personification, which is purely a rhetorical device, with anthropomorphization. I am not anthropomorphizing GPUs. There is nothing particularly weird about personifying model weights, as we do with computer programs, cars
50.
by jchw 2mo ago
This seems to always be explained in a confusing and sometimes wrong manner, so I may as well also try to explain it in a confusing and sometimes wrong manner too. Both Windows and Linux allow applications to map memory in an uncommitted st
51.
by jchw 2mo ago
Personally I wouldn't want bots running autonomously on a repo, even if there were other bots cross checking them. But that having been said, I'd also say that my experience was similar with human code: it is rare to not find at l
52.
by jchw 2mo ago
It's easy to dismiss "taste" when you either have none or just fail to appreciate it, but nothing gives you an appreciation for the importance of taste like LLMs. There is no benchmark for taste, so while many things improve
53.
by jchw 2mo ago
If you said this in 2025 I would've 100% understood, but to be honest getting AI models to do a pretty good job on day-to-day ticket work has become so boring that we don't even bother using the top tier models and higher effort s
54.
by jchw 2mo ago
Although I am not exactly sure what you mean, I am not really claiming it is doing anything I couldn't do - but yes, it does so with much less effort. For example, I can have it set up probes and tracing on Linux that I personally woul
55.
by jchw 2mo ago
To be fair, I actually do run xhigh as my default. However, for the first time in my experience of trying and using LLMs, with Sol.. sometimes I feel confident enough to set the effort level to "Low". I just had Sol prototype so
56.
by jchw 2mo ago
I think Fable's dominance is overstated. It definitely has the lead, but quantifying what that lead actually is is really hard. I'm using GPT 5.6 Sol to do some shit that I personally would consider "crazy" - low level u
57.
by jchw 2mo ago
You can only get a very limited uBlock Origin facsimile, not the real thing. A lot of browsers have a uBlock Lite. It's missing a lot of useful features IMO.
58.
by jchw 2mo ago
Not that I'm aware of. I'm on my phone right now but any KMS driver that supports KMS and dma-bufs should work. The basic idea is to enumerate the CRTCs to get the relevant framebuffer plane resources and then use dma-buf exports
59.
by jchw 2mo ago
Based on the literal hundreds of billions we are pouring in, AI still has a long way to go to prove it is worth what we are putting in. Anyone who thinks it doesn't is delusional. I don't really give a shit if that makes some peop
60.
by jchw 2mo ago
Okay now let's try that again with the whole industry...

More ›