irthomasthomas

6 ms·

61.
by irthomasthomas 1mo ago
glm-5.3, kimi k3 and qwen3.8 are SOTA.
62.
by irthomasthomas 2mo ago
I really don't know. It was built before prompt caching was common, and the switching cost was much lower.
63.
by irthomasthomas 2mo 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.
64.
by irthomasthomas 2mo 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
65.
by irthomasthomas 2mo ago
Openrouter is going to cost you a lot more than the 5% fee, unless you lock the provider.
66.
by irthomasthomas 2mo 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.
67.
by irthomasthomas 2mo ago
Animation was not requested.
68.
by irthomasthomas 2mo 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.
69.
by irthomasthomas 2mo ago
Have you seen the news about decrypting the hidden COT in U.S. models? [0] The decoded logs revealed instances where Claude memorized answers to test questions beforehand while making its final output look like it had derived the answer ste
70.
by irthomasthomas 2mo ago
That is more an indictment of AA than DS
71.
by irthomasthomas 2mo ago
rumor is that is what Ilya has done at SSI.
72.
by irthomasthomas 2mo ago
then what is the point of using operouter for this model? Just use the deepseek API and save the 5% fee on top of the better caching rate.
73.
by irthomasthomas 2mo ago
But they don't perform the same test on other models as far as I could tell? So we don't lnow if this is peculiar to kimi models or not.
74.
by irthomasthomas 2mo ago
From: Che Chang <redacted> Date: Feb 23, 2026, at 8:04 PM Subject: Re: Former Apple Employees at OpenAI Retaining Non-public, Confidential, and Proprietary Information To: <redacted> Hi [Apple in-house legal counsel] and [Apple
75.
by irthomasthomas 2mo ago
I think it's interesting to see them visibly struggling to improve. Claude pelicans aren't much better today than they where 18 months.
76.
by irthomasthomas 2mo ago
I guess expert+chatgpt beats chatgpt alone, so why not hire top experts to drive the search?
77.
by irthomasthomas 2mo ago
Why you think that?
78.
by irthomasthomas 2mo ago
"Senators don't have the luxury that a research scientist has of waiting for evidence." - 1977 https://www.youtube.com/watch?v=xbFQc2kxm9c FYI obesity 1977 ~14% 2026 40%+ T2 Diabetes 1977 ~3% 2026 ~12%
79.
by irthomasthomas 2mo ago
It can still generate a wrong reference and select the wrong verse.
80.
by irthomasthomas 2mo ago
Two world wars and one world cup...
81.
by irthomasthomas 2mo ago
Does x help you avoid that deadspot in the middle, where the flaps cant reach?
82.
by irthomasthomas 2mo ago
But you can weigh up the evidence. A crime has been commited afterall.
83.
by irthomasthomas 2mo ago
interesting... thanks.
84.
by irthomasthomas 2mo ago
But zero evidence provided that this was an unsupervised agent attack. I still find it incredible that a company who protect their IP so much would allow these dangerous experiments to run unsupervised and risk leaking their secrets. Why
85.
by irthomasthomas 2mo ago
If it's true that they run agents like this unsupervised, it is only a matter of time before an openai agent leaks its model weights.
86.
by irthomasthomas 2mo ago
Unless openai release the logs we have only their word that this was done fully autonomously and without their knowledge by an agent running their newest super powerful model. For all we know they could have bought zero days and left them l
87.
by irthomasthomas 2mo ago
No. Lora is for tuning behaviour and how the model applies what it learned in training. Teaching a model new facts is still expensive.
88.
by irthomasthomas 2mo ago
Benchmarks show a large drop in quality as context grows. Your opus will be acting like haiku above 300k tokens.
89.
by irthomasthomas 2mo ago
Yeah something is up. I have the same problem with K3 as I had with earlier kimis. I ask it to write <bash>code</bash> every turn, that does not seem very difficult, but kimi gets this wrong a large percentage of the time.
90.
by irthomasthomas 2mo ago
I use dvorak on PC but qwerty on phone

More ›