Claude is Anthropic’s AI assistant for writing, coding, analysis, and enterprise workflows, with newer tools such as Claude ...

Overview Present-day serverless systems can scale from zero to hundreds of GPUs within seconds to handle unexpected increases ...

Samsung is sunsetting its own chat app while Apple adds end-to-end encryption to its already-live RCS, hinting that the final texting firewall might soon drop. Meanwhile, Cloudflare drafts a ...

Google launches Veo 3.1 Lite, halving AI video generation costs to $0.05 per second at 720p. The model targets developers ...

人工智能可以编写复杂的代码，顺利通过各类高难度考试，给人带来一种机器智慧已经比肩人类的错觉。 ARC-AGI-3正式推出，把当前最聪明的智能体扔进一个没有任何说明书的简单像素游戏里，得分连1%都不到。

凌晨，硅谷又炸了锅。 谷歌DeepMind毫无征兆地甩出Gemma 4全家桶——四款开源模型，从2B到31B，全部基于Gemini 3同源技术打造。这不是一次常规更新，而是对开源大模型格局的彻底重写。 最让同行后背发凉的数字：31B Dense模型，Elo评分1452，开源榜单第三。排它前面的两位，参数分别是它的20倍和30倍。用三十分之一的体量，打出了同等水平的成绩——参数效率这个词，被谷歌重新 ...

Google dropped Gemma 4 on April 2, 2026, and it's a game-changer for anyone building AI. These open models pull smarts straight from Gemini 3, Google's top ...

IQuest团队提出的"代码流"训练范式彻底改变了这种状况。他们不再让AI简单地学习孤立的代码片段，而是让它观察和学习整个软件项目的演化历程。这就像让一个学徒工程师跟着师傅完整地参与一个项目从构思、设计、实现到优化的全过程，而不是只让他背诵零散的技术 ...