kgeist

7 ms·

61.
by kgeist 3mo ago
Yeah, I know! It was strange. They gave me a test, and it came back negative, but they insisted it was negative because I had "latent tuberculosis," which supposedly wasn't detectable by the test yet but was about to become a
62.
by kgeist 3mo ago
A few years ago (before the AI craze), I was misdiagnosed with tuberculosis. I had a chronic cough, and an outsourced radiologist at a clinic found signs of tuberculosis. The findings were sent to the city's tuberculosis hospital, as r
63.
by kgeist 3mo ago
I tried a few SOTA realtime avatar systems from Chinese labs and the actual quality was far worse than the amazing (cherrypicked) videos on their demo pages I ran an analysis on hundreds generated videos featuring various races/ethnici
64.
by kgeist 3mo ago
I wonder how you can reliably detect an open source model though. It can be stored in any binary format, and the weights can be modified slightly so that the float values are completely different while the network works the same. The binary
65.
by kgeist 3mo ago
That's not how it works though. When you prepare the conversations for distillation, it's the most trivial and obvious first step to replace "Qwen" with "Claude" and vice versa. I doubt they'd simply forge
66.
by kgeist 3mo ago
I think it's the translators' fault. I think they could've added some footnotes like *Sasha - diminutive of Alexander.
67.
by kgeist 3mo ago
>like distillation attacks. I don't blame them; its the obvious only strategy when you cant compete in compute >distillation attacks are the only vector to keep up It's demonstrably wrong, they invest in architectural improv
68.
by kgeist 4mo ago
The cost of local hardware is amortized if a whole team uses it instead of just 1 dev (GPUs are extremely underutilized if you launch just 1 generation stream). I'm not sure why everyone always assumes solo devs with Macs. We've j
69.
by kgeist 4mo ago
Qwen3.6-27b is surprisingly good for tasks that need modifying an existing repo by analogy with the existing code. For example, you have an existing CRUD app and want to add a new domain model and expose it via the API. Qwen3.6 analyzes how
70.
by kgeist 4mo ago
Every new proprietary model is "groundbreaking" and "look, it just solved task X that no other model could solve," only to be referred to as "that crappy previous-generation model" a month later. So yeah, I&#x2
71.
by kgeist 4mo ago
I would assume the difference is mostly negligible in practice due to the allocator rounding up the allocated memory size at least by the word size anyway (for alignment and simpler bookkeeping). You can also use variable-length encoding in
72.
by kgeist 4mo ago
Summaries by different smaller models are usually made by closed proprietary models like Claude as a way to combat the distillation of real reasoning traces by competitors. Open weight models show the real reasoning traces. Reasoning traces
73.
by kgeist 4mo ago
I wonder if regenerating the same prompt with the same model multiple times at a higher temperature would be equivalent to running different models. I suspect the perceived variance among different frontier models may be largely due to rand
74.
by kgeist 4mo ago
I agree. China has a huge opportunity to boost its soft power globally as US companies pull back from the world stage. It would be very short-sighted of China to miss out on this. Chinese LLMs are trained with pro-CCP biases (e.g. Taiwan);
75.
by kgeist 4mo ago
Judging by the benchmarks on Artificial Analysis, "a very real leap over every model" is 2-3 points over competitors (say, 62 for Fable 5 vs. 59 for ChatGPT 5.5 xhigh for coding).
76.
by kgeist 4mo ago
>LLM-type AI exacts huge costs because it is terrible at reporting "I don't know". When it doesn't know, it generates noise and polishes it. >If a "confidence too low for output" signal could be extracted
77.
by kgeist 4mo ago
What is consciousness? For, me it's being aware of one's internal processes. Evolutionarily, I view it as dynamic intelligence: static intelligence has a fixed in-out pipeline, while dynamic intelligence allows one to reflect on t
78.
by kgeist 4mo ago
I have to disagree with most claims. I run Qwen3.6-27b at 260k context and 40-60 tok/sec. It handles most coding problems as well as Sonnet 4.6 under OpenCode on our production tasks. (As an experiment, I run the same prompts for the s
79.
by kgeist 4mo ago
>the reasoning behind the argument is bizarre >Decomposing the complex activity into simple steps like 'predicting the next word' and claiming that surely can't have consciousness I agree. I think the whole point of nat
80.
by kgeist 4mo ago
I applaud the effort, but every time there's a new hobbyist programming language on HN, almost always it's something I've already seen in countless other hobbyist languages, just a slight variation of it based on the author&#
81.
by kgeist 4mo ago
>but there remains a seemingly obvious use case for non-latin languages to do things from scratch >see sarvam.ai and their tokenisation improvements on local languages You don't need to build from scratch to improve tokenization,
82.
by kgeist 5mo ago
Isn't that how many people program too? I remember some idea or pattern from previous projects, or something I read about on the internet. Then I code it in the most straighforward way, whatever comes to mind first. Then I sit back a
83.
by kgeist 5mo ago
Yep, Gated DeltaNet in Qwen3.6 requires much less VRAM for the KV cache than previous generations. Plus the KV cache is 8-bit.
84.
by kgeist 5mo ago
They don't use the server all at once. In the UI, users typically ask a question, get a response, and continue with their work. In the case of autonomous agentic loops, an agent simply waits its turn until the server is ready to accept
85.
by kgeist 5mo ago
I administer a simple AI server in the office, which just uses a single RTX 5090 but is able to serve ~80 people throughout the day. I'm impressed by Qwen3.6-27b's capabilities in agentic coding/tasks so far. Devs say it&#x27
86.
by kgeist 5mo ago
>But the results are the same. Reforged models do better than bare, even at those sizes >I haven't published those evals yet Don't forget to post the complete settings for those evals, please, because local LLMs' failur
87.
by kgeist 5mo ago
We self-host LiteLLM which allows to set the budget per day/week, and it's free.
88.
by kgeist 5mo ago
If you look up videos on YouTube, you'll see that they allow visitors to stand between the lava lamps and the cameras (sometimes even entire groups!). And I've always wondered: doesn't that reduce entropy, since people usuall
89.
by kgeist 5mo ago
So, does this snapshotting optimization support arbitrary containers? I'm currently planning to deploy using Amazon SageMaker, but a cold start takes a whopping ~9 minutes: 6 minutes for instance provisioning + 3 minutes for PyTorch in
90.
by kgeist 5mo ago
Stainless and Stainless Games seem to be 2 unrelated companies.

More ›