woadwarrior01

7 ms·

61.
by woadwarrior01 2mo ago
Evidently, Claude's tokenizer vocabulary size is ~15k[1]. On one hand, it's quite mind blowing. On the other hand, Anthropic models' token (in)efficiency makes a lot of sense in that light. [1]: https://xcancel.com
62.
by woadwarrior01 2mo ago
This is so reminiscent of Cider (Google's internal cloud IDE) which had the best multiplayer support that I've ever seen. It was such a joy to pair program with a colleague on a different continent.
63.
by woadwarrior01 2mo ago
All this posturing is all for naught, because people can do proper logit distillation into smaller models with open-weight models.
64.
by woadwarrior01 2mo ago
TBF, he has done more than his fair share of time working in BigTech.
65.
by woadwarrior01 2mo ago
Good luck writing fanfics with primitive top-p and top-k samplers.
66.
by woadwarrior01 2mo ago
Huggingface now owns llama.cpp, btw. https://huggingface.co/blog/ggml-joins-hf
67.
by woadwarrior01 2mo ago
ICYMI, llama.cpp was also VC funded. Search for "ggml" on this page: https://aigrant.com
68.
by woadwarrior01 2mo ago
For once. A former colleague once quipped: The only way to make decent money in the business of language runtimes is to get a job at a big company that depends on it.
69.
by woadwarrior01 2mo ago
The value has already been materialized. They were acquired by Qualcomm for $3.9 billion.
70.
by woadwarrior01 2mo ago
IIUC, it's the ability to export layered PSD files with images from multiple sensors.
71.
by woadwarrior01 2mo ago
When you get around to doing that, you might find the Github repo and its accompanying medium article useful. https://github.com/0Itsuki0/SwiftUI_P2PWithNetwork
72.
by woadwarrior01 2mo ago
There's nothing to copy. Apple's iOS Camera app already uses multi-camera image fusion in many cases. This app builds on the Multi-Camera AVCapture feature that Apple opened up in iOS 13. https://developer.apple.com&#x2
73.
by woadwarrior01 2mo ago
The Claudish in the blogpost makes it really hard to ready. Also, TinyLlama 1.1B lol.
74.
by woadwarrior01 2mo ago
> Note how they're much smaller than all other models in the comparison yet match or exceed them. There's a strong incentive to cherry pick in self-reported comparisons. If there is a model that's better, it gets left out.
75.
by woadwarrior01 2mo ago
Better yet, use Apple's container CLI if you're on a Mac, instead of the docker bloatware. container run --rm -it -v "$(pwd)":/work -w /work myaiimage /bin/bash
76.
by woadwarrior01 2mo ago
FWIW, I left a VC backed startup where I was an equal co-founder because my co-founder was committing securities fraud (blatant lying about customers and traction in investor update emails, amongst other things). I presented all the evidenc
77.
by woadwarrior01 2mo ago
Something related that I've thought of for a long time. Browsers have had the BroadcastChannel API for almost a decade now, and yet sites like HN and Reddit don't use it to synchronize state between multiple open tabs.
78.
by woadwarrior01 2mo ago
I suspect this acquisition was a response to NVIDIA's acquisition of Groq. Cerebras is already public. AFAICT, there are 8 other startups in the space, some of which have mature products: Etched, d-Matrix, SambaNova, Tenstorrent, Posit
79.
by woadwarrior01 2mo ago
Haven't you seen the kernel optimization case study at the bottom of the page? They compare against GPT-5.6 Sol and their model is worse.
80.
by woadwarrior01 2mo ago
I haven't compared it for perf, but at first glance this appears to be an LLM (text-only) while both binary and ternary bonsai 27B models are VLMs (where the vision tower and the adapter MLP weights are unquantized and kept in float16)
81.
by woadwarrior01 2mo ago
At first glance, this work might also be relevant for quantizing diffusion transformers (something I'm working on at the moment).
82.
by woadwarrior01 2mo ago
Wow, I can't believe that the author is still updating the book! I'd learnt protected mode assembly from an older version of this book, decades ago. And IIRC, there was an even older 16-bit version of the book back then.
83.
by woadwarrior01 2mo ago
Deuterium or tritium? Deuterium is a stable isotope and does not decay.
84.
by woadwarrior01 2mo ago
I'm not making any moral judgements based on it, merely observing it.
85.
by woadwarrior01 2mo ago
> The measured result is a reference point, not a performance ceiling. Claude was here.
86.
by woadwarrior01 2mo ago
Could this possible be the first AI worm? Or are there any priors to this?
87.
by woadwarrior01 2mo ago
Wouldn’t that be TAT?
88.
by woadwarrior01 2mo ago
Those were voluntary disclosures by two Anthropic researchers and the security firm Calif. I know one more CVE on the list that was discovered using an AI agent and wasn't disclosed as such. I suspect there are many more.
89.
by woadwarrior01 2mo ago
Interestingly enough, this and many other related ethical conundrum were covered in the Nuremberg Charter from 1945.
90.
by woadwarrior01 2mo ago
Hey, thanks for doing the right thing. There's no need for an apology.

More ›