分类： AI 资讯

Models like Kimi K3, Grok 4.5, and Muse 1.1 may prevent the dominance of 2-3 frontier labs with 90% inference margins from hurting other AI ecosystem layers (Gavin Baker/@gavinsbaker)

Models like Kimi K3, Grok 4.5, and Muse 1.1 may prevent the dominance of 2-3 frontier labs with 90% inference margins from hurting other AI ecosystem layers (Gavin Baker/@gavinsbaker)

知名投资人 Gavin Baker 指出，Kimi K3、Grok 4.5、Muse 1.1 等新一代模型正在涌现，它们有望阻止少数前沿实验室凭借高达 90% 的推理利润统治行业，避免对其他 AI 生态层造成损害。这暗示了市场格局可能从“垄断式”向“多元化”演变。

‘Whoever came up with this is a massive idiot’: LG’s gaming monitors and TVs are facing a user revolt, due to seemingly installing adware on PCs — and telling you to warn guests they may be recorded by AI features, to comply with ‘wiretapping’ laws

'Whoever came up with this is a massive idiot': LG's gaming monitors and TVs are facing a user revolt, due to seemingly installing adware on PCs — and telling you to warn guests they may be recorded by AI features, to comply with 'wiretapping' laws

LG 的部分游戏显示器被发现未经用户许可自动安装广告软件，同时其智能电视新用户条款要求使用者必须提前告知访客“可能被 AI 语音功能录音”，否则需自行承担“窃听法”合规责任。两件事叠加引爆了用户不满。

Anthropic 工程师代码产出提升 8 倍：背后的上下文工程系统

Anthropic 工程师代码产出提升 8 倍：背后的上下文工程系统

Anthropic 内部的工程师团队在没有更换模型、硬件或扩招人员的情况下，每日代码合并量提升至一年前的 8 倍。核心变化在于他们不再将精力花在打磨提示词上，而是系统性地构建 AI 智能体运行所需的“上下文”信息环境，这一方法被命名为“上下文工程”，正在取代提示词工程成为 AI 开发的新范式。