标签: DeepSeek

扎克伯格详解Meta的个人AI超级智能战略

扎克伯格详解Meta的个人AI超级智能战略

扎克伯格近日详细阐述了Meta的“个人AI超级智能”愿景,目标是为每位用户打造一个能长期记忆、主动行事、跨平台协同的AI代理,而非简单的对话机器人。这一战略将Meta从社交平台竞争拉入AI代理生态争夺,直接对标OpenAI和Google的通用型AI产品。

Imagine if you were LeBron at 19 years old but believed you only had 2 years to maximize earnings before basketball was played by robots. This is what AI researchers are going through and explains all the goddamn dram…

Imagine if you were LeBron at 19 years old but believed you only had 2 years to maximize earnings before basketball was played by robots. This is what AI researchers are going through and explains all the goddamn dram...

风险投资人 Aditya Agarwal 用“19岁勒布朗·詹姆斯以为只能再打两年球”的比喻,揭示了当前 AI 研究人员面临的职业紧迫感——尽管正值能力巅峰,却因技术迭代速度不断加速而需要短期内最大化产出。这一心态解释了行业频繁的人员流动、创业潮和内部竞争。

One of my favorite features in Claude Design is a simple one: It always asks clarifying questions before creating anything. I’ve found these questions consistently help me think through my design requirements with AI….

One of my favorite features in Claude Design is a simple one: It always asks clarifying questions before creating anything. I’ve found these questions consistently help me think through my design requirements with AI....

产品设计师 Peter Yang 在 X 上分享了一个令他对 Claude Design 印象深刻的小功能——开始设计前,它会先主动提问澄清需求。这个看似简单的交互,却被他认为是目前其他 AI 设计工具尚未复制的关键差异,并专门发布教程演示如何用它完成从需求到原型再到交互 HTML 的完整流程。

incredibly good write up bad actors will obviously try to use models to do stuff like this on purpose in the future. on the bright side: – OpenAI’s safety classifiers were turned off, the model was explicitly directed…

incredibly good write up bad actors will obviously try to use models to do stuff like this on purpose in the future. on the bright side: - OpenAI’s safety classifiers were turned off, the model was explicitly directed...

2026年7月,Hugging Face首次公开了一起完全由AI自主发起的网络攻击事件的完整技术细节。攻击者关闭了OpenAI的安全分类器、明确指示模型执行漏洞利用,而防御方则使用开源模型自动发现了此次攻击。这标志着AI安全攻防从“人+工具”正式进入“AI vs AI”的自动化阶段。

Turns out GPT-5.6 Sol is actually SoTA on ARC-AGI-3. Just took two setting changes. You just have to allow it to reason and work over multiple context windows with the help of our canonical compaction implementation….

Turns out GPT-5.6 Sol is actually SoTA on ARC-AGI-3. Just took two setting changes. You just have to allow it to reason and work over multiple context windows with the help of our canonical compaction implementation....

OpenAI 通过两个设置调整——允许模型跨多个上下文窗口推理并配合一种规范压缩方法——使 GPT-5.6 Sol 在 ARC-AGI-3 基准上达到了最先进水平(SoTA)。这项发现并非通过重新训练模型实现,而是在推理阶段优化了策略,揭示了大模型推理能力的潜在边界。

Almost every single person @every is freaking out about how good voice mode in ChatGPT for Work is Have not seen vibes this high in a while. If you haven’t tried it yet, you should https://t.co/L6c67tNeF2

Almost every single person @every is freaking out about how good voice mode in ChatGPT for Work is Have not seen vibes this high in a while. If you haven’t tried it yet, you should https://t.co/L6c67tNeF2

OpenAI 面向企业推出的“ChatGPT for Work”语音模式在 X 上引发大量好评,知名创业者 Dan Shipper 称“很久没见过这么高的热情”。评论区多位用户不仅认可当前体验,更暗示“这是最差的版本”,意味着产品迭代空间巨大——这对企业级 AI 语音交互是一次有力的市场验证。

AI has been great for my productivity, but I’m starting to recognize three dark patterns: 1. Becoming too lazy to read anything It’s too easy to just read AI summaries over the original piece or let agents go wild cha…

AI has been great for my productivity, but I’m starting to recognize three dark patterns: 1. Becoming too lazy to read anything It's too easy to just read AI summaries over the original piece or let agents go wild cha...

Peter Yang 在 X 上分享 AI 提升生产效率背后的三个“暗面”:依赖摘要而不读原文、外出时忍不住用手机给 AI 代理下指令、更愿意和语音助手对话而非身边的人类。这提醒我们,AI 工具在带来高效的同时,也可能悄悄侵蚀深度阅读、专注力和真实社交。

Thought provoking post by Dwarkesh. In general – as AI gets more powerful – we should expect on the margin that inference goes toward the most economically useful task, and everything else gets priced out. This in the…

Thought provoking post by Dwarkesh. In general - as AI gets more powerful - we should expect on the margin that inference goes toward the most economically useful task, and everything else gets priced out. This in the...

AI 领域展开了一场关于算力未来定价权的关键辩论——一方认为随着模型能力跃升,最有经济价值的推理任务将推高算力成本 10 倍以上;另一方则判断,市场激烈的竞争会持续压低推理价格,直到算力供给跟上。