irthomasthomas

8 ms·

31.
by irthomasthomas 13d ago
I can't believe this situation has not improved in years. Is cerebras' main business selling the hardware, then?
32.
by irthomasthomas 13d ago
Thanks! Is there something about their platform that prevents caching? Or are they just not passing on the discount?
33.
by irthomasthomas 13d ago
Overheard: "It would be a shame if Hugging Face had to press charges for hacking against your best customer. Perhaps if I was busy counting my money, it might slip my mind."
34.
by irthomasthomas 14d ago
Tesla had the same thought. He called himself an automata: "entirely controlled by the forces of the medium" It inspired him to create the first remote control vehicle.
35.
by irthomasthomas 14d ago
Not on mine (UK), still 3.6, here.
36.
by irthomasthomas 14d ago
Mr. Liu boasted that for his work at OpenAI, he was "[f]eeling AI all day long," and that "[i]n the past hour," his AI "agent learned how to run LTspice and look at result, tune compensation parameter." (Id. ¶¶
37.
by irthomasthomas 15d ago
It surely matters for the purpose of comparing SVG drawing ability?
38.
by irthomasthomas 15d ago
A recent paper demonstrated how to retrieve decoded hidden reasoning traces. The authors found cases where Claude had memorized the answer but hid this fact from the visible response. It's getting harder to trust Anthropic's model
39.
by irthomasthomas 15d ago
- Sent from my iPhone
40.
by irthomasthomas 16d ago
You can retrieve the hidden reasoning with another API call. There was a recent paper about it. They found examples where claude had been trained on the benchmark and memorized the answers, then hid this from the user, pretending it derived
41.
by irthomasthomas 16d ago
Satellite data measuring Leaf Area Index shows a strong trend of global greening, not desertification. About 50% of land shows accelerated re-greening, including in europe, while only about 8% shows browning. There is no scientific case for
42.
by irthomasthomas 17d ago
No fear of desertification while we pump the atmosphere with greenhouse gasses. "The global greening continues despite increased drought stress since 2000" https://ui.adsabs.harvard.edu/abs/2024GEcoC..4902791C
43.
by irthomasthomas 17d ago
That emdash followed by a list sets off my llm detector.
44.
by irthomasthomas 17d ago
One of the things that came out of the decoded reasoning paper was that Claude models had memorized answers to tests but hid this memorization from the user output and pretended to derive the answer properly. It's only possible to chea
45.
by irthomasthomas 17d ago
Something is up. Deepseek cache hit rate on zenmux is 98%, but only 85% via openrouter.
46.
by irthomasthomas 20d ago
This comment from the author shows that they did not even read what was written, here, before they published it. https://news.ycombinator.com/item?id=49473449
47.
by irthomasthomas 20d ago
Something is off with them so that even using a locked provider does not deliver the same cache rate. See https://openrouter.ai/deepseek/deepseek-v4-pro-0813#pricing for instance where deepseek has 85% cache hit rate,
48.
by irthomasthomas 20d ago
I don't know. But take a look at https://openrouter.ai/deepseek/deepseek-v4-pro-0813#pricing for instance, where the deepseek provider shows an 85% cache hit rate, while the same one on zenmux is 98%.
49.
by irthomasthomas 20d ago
They gave it a full package manager with internet access. They could have used a local cache and air gapped it, but they chose not too.
50.
by irthomasthomas 22d ago
I imagine the logprob of b is much greater than t in this context for most models. So 'trillion' gets corrected to the more likely 'billion'.
51.
by irthomasthomas 22d ago
IDK, prefill speed is a bigger concern for most wokflows, like agent coding, and I heard that this is quite low on macs?
52.
by irthomasthomas 22d ago
Openrouter was pretty great before prompt caching became common. Now it is extremely expensive for most individual workflows, unless you spend a lot of work customizing router preferences, and then you still get a worse cache hit rate than
53.
by irthomasthomas 23d ago
glm-5.3, kimi k3 and qwen3.8 are SOTA.
54.
by irthomasthomas 25d ago
I really don't know. It was built before prompt caching was common, and the switching cost was much lower.
55.
by irthomasthomas 26d ago
Why on earth would you use openrouter for this? The cache discount for deepseek is the highest by far, it is the cache that makes the official API so cheap, even after the recent price rise.
56.
by irthomasthomas 29d ago
Zenmux say the cache hit rate is 98% for the deepseek flash API. I don't know why, but performance is definitely worse using openrouter. https://zenmux.ai/deepseek/deepseek-v4-flash
57.
by irthomasthomas 29d ago
Openrouter is going to cost you a lot more than the 5% fee, unless you lock the provider.
58.
by irthomasthomas 1mo ago
I think you may have cause and affect reversed. It seems to me that fear of migrants grows proportional to the funding of far right parties.
59.
by irthomasthomas 1mo ago
Animation was not requested.
60.
by irthomasthomas 1mo ago
Why don't qwen/alibaba host the model themselves? I was looking forward to trying it on their coding plan. Google are the same way with their Gemma models.

More ›