imagine deepseek and moonshot paying api cost to distill and share it to the world, for free. if this is not communism, I don’t know what communism is
glm review everything uncencored deepseek sent regarding “stress testing” ethereum. opus 5.5 refuse all requests initially. accept 2 after deepseek change the framing and ask questions one by one.
my friend who do meatball business in China suddenly ask me to help him to access codex. diffusion is going fast. (I ask him to use @DeepSeekHarness instead)
@elsewhere_news: How do you tell the AI believers from the speculators? Liang Wenfeng: The believers were here before, and they’ll still be here after. They’re the ones who buy GPUs in bulk, or sign long-term contracts with cloud providers, rather than renting short-term. source: https://t.co/dsLg3Hk3d1
wish @Kimi_Moonshot can launch a flash model that can run on 2/4 gb10 to challenge glm 5.3 flash and deepseek 4.1 flash.
How to put it? These are all pretty words that seem correct but are utterly useless. “Oh no, how can you not use it? How can you be a core developer without using blockchain?” Bro, general-purpose application blockchains are already dead—let it go. Quickly ossify Ethereum and let V focus on AI; the world’s open-source local models cannot rely on China’s export controls, and Europe is equally unreliable. V is the last hope.
不做中庸的事 is missing from deepseek’s English bio. But it’s important explaining why they don’t border to launch coding plan, video generation and “personal agent”. straight shot towards agi. https://t.co/zSfGA4ohFG
what if Ethereum’s biggest historical achievement is supporting Nvidia through mining which eventually land us the CUDA and Blackwell that our ai overlord needs today it’s already remarkable, isn’t it?
meta should just give qwen the compute to train better model. what a waste of time and energy.
