星辰大海

Google DeepMind 刚发了 Gemma 4，这一代其实挺实用的。 我觉得重点不在“更强”，而在于它的定位很清楚： 1/ E2B / E4B：更轻，资源占用低，适合手机和边缘设备 2/ 26B / 31B：能力更完整，适合在 PC / 本地环境跑 3/ 多语言支持（140+） 4/

when you consumed a double shot espresso, a bottle of passionfruit celcius, and 4 magic mind before 10am in the office on Monday

盗梦空间 既视感

x.com/i/article/2042…

Canada is cooked

From concept to testnet. Come build.

用了一天的 Opus 4.7，你注意到了吗？Claude 再也不拿那种便宜的 Haiku 模型来糊弄我了。今天的每一个问题都是实打实地用 4.7 来回答，勤勤恳恳的，值得表扬

谢谢分享，我同时也推荐可以看一下下面这个 podcast youtube.com/live/pwaWilO_P…

💯

Could you please fix this annoying bug, or consider addressing it before release? The new updates notification blocks the content I’m reading when I use J or K to move to the previous or next post. Nikita Bier

asking for a fren, how to break this vicious circle?

TIL: “Chai” means tea in Hindi and Urdu.

The next bottleneck in AI isn’t model performance. It’s data. From scraped internet content → real-world, consented, high-context data.

TGIF in case you don’t know this ai means love in Chinese

The only thing better looking than the 3rd guy on the left is what we’re releasing next week Here’s the squad from our latest sprint 💪🏼

stay hungry, stay foolish