Kayou

Joined 13 years ago

Karma

121

Submissions

87

Member since

Aug 29

View on Hacker News

Recent submissions

Comment · 2 months ago

Why do all LLMs do a particle simulation when you ask them this prompt ? Qwen3.6, Qwen3.8 and Ling 3.0 Tiny all did the same thing !

I find Ling 3.0 tiny particularly interesting as it looks really nice for a tiny model with 7.9B total parameters, with only 1.3B parameters activated per token. Here is the result https://coolthing-ling-3-tiny.tiiny.site (sorry for the weird hosting, first I found that worked)

(it cost me almost 0 cents and done in 49 seconds)

Comment · 2 months ago

The dense 27b Qwen on M4 Pro has a prompt processing speed of around 125tok/s which makes it ok to ask a quick question but impossible to use in an agent, as processing the first prompt of the agent with the tools and instruction can easily be 10 000 tokens

In this case the 35b a3b makes sense as it has a PP speed of around 800tok/s

Comment · 4 months ago

I agree with OP that the name, while maybe not deliberately, is really confusing in 2026. I thought that was a wind turbines or a solar panel technology, definitely not gas.

Comment · 8 months ago

Wait, the Q4 quantization which is more than 20GB fits in your 16GB GPU ? I didn't know that was possible, I was always restricting myself to smaller model than the VRAM I had

Comment · 1 year ago

If you're going to use a lib like pLimit to limit the concurrency of a map, just use pMap which accepts an argument to limit the concurrency and makes it more readable and straightforward

https://github.com/sindresorhus/p-map?tab=readme-ov-file#usa...

Comment · 2 years ago

I have thrown people under the bus while under pressure from my manager to explain why reviewing a Pull Request takes me several hours or days while other members did it in a few minutes. It was the wrong thing to do but the pressure was unbearable.

Comment · 4 years ago

Thanks for the article on energy storage with hydrogen to resolve peaks on the grid. Seems like a nice solution on paper, although ineficient according to this article citing someone from the MIT, round-trip efficiency of 18%-46%.

https://www.spglobal.com/marketintelligence/en/news-insights...

Comment · 4 years ago

In that case what prevents us from building energy storage facilities that produce hydrogen at night and release the electricity back into the grid during the day? Wouldn't it be more efficient than transporting the hydrogen by truck and transporting the hydrogen battery on every train?

Comment · 4 years ago

Same on MacOS and Firefox with a 4K screen ans same on Firefox and a Pixel 6, scrolling lags hard on the section with pictures. Making a bad first impression of this new OS.

Comment · 4 years ago

What lense do you have to take your picture ? You might want to zoom a bit to make it look like you see it, at least a 100mm equivalent I would say.

Comment · 5 years ago

Yeah, back in the day I would always be afraid to fry my device when plugging a 3rd party charger, because of the polarity or voltage. Now worst thing that can happen is... nothing. For me that's an improvement.

Comment · 5 years ago

Inner city Paris is limited to 30km/h, I don't think going 25 km/h would be a problem.

Comment · 5 years ago

Sure enough for the power consumption of the modems, but I don't think the 5G networks are solar powered in Australia. Using multiple networks at the same time will use more power than using only one.

Comment · 5 years ago

Yeah, and I think it really shows that WhatsApp, which wasn't developped by Facebook originally, doesn't benefits from the compression. The app seems very optimized originally, but maybe I'm wrong.

Comment · 6 years ago

Yes, and don't forget to shutdown Discord too! They will never be able to use another tool to organize these raids, this is genius. (this is sarcasm, of course)