yogthos

11 ms·

271.
by yogthos 4mo ago
That's the fundamental difference with open models, anybody can run and tune them any way they like. The real difference in philosophy is that Americans companies treat the model as the product, while Chinese companies see models at in
272.
by yogthos 4mo ago
I'm waiting for that to happen as well since the price difference makes it very difficult for companies like Anthropic and OpenAI to compete. And we already have precedent for this with stuff like EVs, phones, and so on. As soon as Chi
273.
by yogthos 4mo ago
This is precisely why I expect that Chinese open models are going to win in the long run. The capability difference isn't dramatic in the grand scheme of things, but the fact that you can run your own is a huge selling point. Even if y
274.
by yogthos 4mo ago
Not if Chinese companies have anything to say about it.
275.
by yogthos 4mo ago
A fantastic move to ensure the rest of the world keeps using Chinese models.
276.
by yogthos 4mo ago
Nobody with even a modicum of understanding of how LLMs work believes any of this. These 'serious thinkers' are just grifters preying on the feeble minded.
277.
Chinese startup claims photonic chip production without DUV lithography(tomshardware.com)
4 points by yogthos 4mo ago|0 comments
278.
by yogthos 4mo ago
neat
279.
Open Reproduction of DeepSeek-R1(github.com)
246 points by yogthos 4mo ago|18 comments
280.
Flatiron is a fast columnar analytics library for Clojure(github.com)
66 points by yogthos 4mo ago|0 comments
281.
Attention Residuals(arxiv.org)
3 points by yogthos 4mo ago|0 comments
282.
by yogthos 4mo ago
I think of it as a genetic algorithm loop. The LLM is basically a mutator function within the loop. If you can define the end shape you're looking for using tests and specification then you can throw the LLM at the problem and have it
283.
The difference in perspectives between superpowers is shaping the race for AI(nytimes.com)
3 points by yogthos 4mo ago|0 comments
284.
by yogthos 4mo ago
I find the trick I use is to get the model to come up with a phased plan, and review it. If I spot anything that seems dumb, I give direction on the way it should be done. And once you finalize that, the model can run through the steps fair
285.
by yogthos 5mo ago
I'm really glad we have an open model that's competitive with the closed frontier ones. This tech is way too important for a handful of corps to decide on how these models are trained and used.
286.
A theoretical reconstruction of the Mythos architecture from first principles(github.com)
6 points by yogthos 5mo ago|0 comments
287.
DeepSeek V4 Pro beats GPT-5.5 Pro on precision(runtimewire.com)
397 points by yogthos 5mo ago|225 comments
288.
Huawei executive credits bans for accelerating domestic chip independence(techradar.com)
22 points by yogthos 5mo ago|2 comments
289.
Data centers consumed 264B gallons of water as drought hits nearly 63% of US(barchart.com)
24 points by yogthos 5mo ago|27 comments
290.
by yogthos 5mo ago
Yeah, that's a good idea. I haven't really been rigorous with tracking the token usage metrics when I started. I was thinking I could compare solving tasks with opencode too and track metrics for both.
291.
by yogthos 5mo ago
I think it's almost certain that we'll be moving to running local models as a default in a few years. The quality of small models has been improving at an astonishing rate in my opinion. My favorite example is how Qwen3.6-27B that
292.
Multi-Robot Cooperative Spatial Reasoning with Multimodal Large Language Models(arxiv.org)
3 points by yogthos 5mo ago|0 comments
293.
Google is quietly laying off staff in its cloud division(businessinsider.com)
19 points by yogthos 5mo ago|1 comments
294.
New IronWorm malware hits 36 packages in NPM supply-chain attack(bleepingcomputer.com)
22 points by yogthos 5mo ago|0 comments
295.
by yogthos 5mo ago
The key problem is that we don't really have a clear definition of what constitutes consciousness. And without having a clear theory of consciousness, it's not really possible to say whether something is conscious or not. Personal
296.
by yogthos 5mo ago
I'd argue that his most convincing argument is in the latter part where he shows that Anthropic doesn't really take the idea that Claude might be conscious seriously either. Because if they did, their constitution would provide it
297.
by yogthos 5mo ago
I disagree that it undercuts the point in any way. The original story didn't appear out of a vacuum, the human consciousness that wrote the story was itself borrowing extremely heavily from the sum total of human ideas. It was shaped b
298.
by yogthos 5mo ago
All of that combined is peanuts compared to what's happening in China. Not to mention that all the panels and most of the wind turbines are produced in China. It's not just a question of installing them, it's having the indus
299.
by yogthos 5mo ago
China having managed to position itself as the main driver of the green transition by investing into key industries illustrates the power of state planning. The markets simply can't operate on horizons of decades because there is no im
300.
Tracing Rays with Jank(jank-lang.org)
46 points by yogthos 5mo ago|0 comments

More ›