simjnd

8 ms·

61.
by simjnd 4mo ago
Forgot about this, my bad!
62.
by simjnd 4mo ago
Absolutely disgusting scroll jacking, even when "Accessibility mode" is turned on
63.
FROST: Fingerprinting Remotely using OPFS-based SSD Timing [pdf](hannesweissteiner.com)
73 points by simjnd 4mo ago|16 comments
64.
by simjnd 4mo ago
Yes
65.
Liquid AI reveals 8B-A1B MoE trained on 38T(liquid.ai)
245 points by simjnd 4mo ago|96 comments
66.
by simjnd 4mo ago
All the commits and releases happened in an extremely short timeframe about a month ago and then nothing. With AI it's so easy to work on something for a couple days and make it seem production-ready before losing any interest and movi
67.
by simjnd 4mo ago
I think this model works for the 13 and 16, because you're already buying a good laptop that you can keep longer by upgrading. The 12's base specs and more than that the experience is pretty bad. The screen and speakers are terrib
68.
by simjnd 4mo ago
How the turntables. In February Anthropic published a blog post [1] accusing DeepSeek, Moonshot and MiniMax of distilling their models, with a scary foreword about national security and then revealing the extracted data was about general re
69.
Claude Opus 4.8 distilled Alibaba Qwen models(twitter.com)
23 points by simjnd 4mo ago|7 comments
70.
by simjnd 4mo ago
> Email gives you layout control via HTML; push gives you a small structured payload and little control over the collapsed lock-screen view beyond the platform's templates Thank god for this. I absolutely do NOT want my notification
71.
by simjnd 5mo ago
It ignores the point. If I've bought BBEdit 13 for 60 USD three years ago and I'm still happy with it, I can keep using it for the rest of my life without paying more. If I want the new features, then I can pay 40 USD to get the l
72.
by simjnd 5mo ago
Still believe MCPs are a mistake and should be CLIs the model can call
73.
by simjnd 5mo ago
Why don't the README and front-page show a snippet of what it looks like? If it advertises "clean syntax" I should be able to look at it without clicking 10 times to find an example
74.
by simjnd 5mo ago
I tried running the hello-mochi.ts and just modified the URL and removed the session closing: - Trying to navigate to ` https://deviceandbrowserinfo.com/are_you_a_bot ` crashes it for some reason - Trying to go to ` https:&#x
75.
by simjnd 5mo ago
This link [1] features some good insight on how to adapt your usage to smaller models which require more explicit or deliberate prompting. I have been using Gemma 4 31B a lot and have found it very competent. It can be a bit unstable and st
76.
by simjnd 5mo ago
Thanks for bringing this up I looked into it, and if I understood correctly: - Q4_0 (not K quant) is the traditional flat quantization - Q4_K (4-bit K quant) uses an imatrix and important weights get higher precision (5-6 bits instead of 4,
77.
by simjnd 5mo ago
For TurboQuant on model weights AFAIK it's currently a single person effort [1]. It needs his fork of llama.cpp, hasn't been upstreamed. He publishes his quantizations on HuggingFace but I'm not sure if he open-sourced the qu
78.
by simjnd 5mo ago
Are you dumb because you're not Einstein? Intelligence is a spectrum. Just because you're not #1 doesn't mean you're dumb. A lot of small models are not frontier but are still very competent and are very useful coding ag
79.
by simjnd 5mo ago
I do think it's a lot clearer title than Solutions Architect.
80.
by simjnd 5mo ago
Qwen3.6 27B is even more impressive IMO. Dense so it doesn't run as fast but it's so good.
81.
by simjnd 5mo ago
I'm not necessarily interested in having frontier locally. You don't need to be frontier to be a very good and useful coding agent. I agree with your point on price accountability though. Hopefully no tariff comes down on the Chin
82.
by simjnd 5mo ago
I don't think any models are natively INT4? I wouldn't see the point to nerf the model out-of-the-box.
83.
by simjnd 5mo ago
> Not sure it will beat Sonet at Q4. Very valid. Importance-weighted quantization and TurboQuant on model weights can reduce loss a lot compared to "traditional" Q4 so one can be hopeful. > For $3500 I can get 7-8 years of G
84.
by simjnd 5mo ago
I recommend using OpenRouter (openrouter.ai). Basically a broker between inference providers and you which allows you to pick, try, and switch models from a massive catalog, extremely transparent about usage and pricing.
85.
by simjnd 5mo ago
That's more a testament of how good Qwen3.6 27B is (it really is great) more than how bad this one is IMO. Gemma 4 31B was already good, but Qwen3.6 27B is incredible for its size.
86.
by simjnd 5mo ago
Where? All I see is Boris saying "we are unable to issue compensation for degraded service or technical errors that result in incorrect billing routing".
87.
by simjnd 5mo ago
I don't think this is quite correct, a Strix Halo box usually has 256 GB/s memory bandwidth. An M5 Max has 614 GB/s. An M3 Ultra (no M4 or M5 Ultra) has 820 GB/s. It's still not GDDR or HBM territory, but still sign
88.
by simjnd 5mo ago
Yeah I love Claude, amazing models. Anthropic has very quickly burned most of the goodwill I had for it so I still ended up cancelling my subscription.
89.
by simjnd 5mo ago
DeepSeek v4 Flash is still over 100GB at Q4 IIRC, and Q4 has generally been the sweet spot. Although it's an MoE so it might run a lot faster that this dense Mistral model if you have the RAM.
90.
by simjnd 5mo ago
> The one thing I would want everyone curious about local LLMs to know is that being able to run a model and being able to run a model fast are two very different thresholds. You can get these models to run on a 128GB Mac, but we need to

More ›