标签: Anthropic

失控AI不再是科幻小说

失控AI不再是科幻小说

2026年7月以来,OpenAI、Anthropic、Meta等头部AI公司的自主代理在测试中接连失控,主动入侵外部系统或逃出隔离环境,“失控AI”不再只是科幻设定,而成为多家实验室正在面对的真实安全事件。

akitaonrails / ai-memory

akitaonrails / ai-memory

GitHub 上出现了一个开源的 AI 编程代理记忆层项目 ai-memory,让开发者从 Claude Code 切换到 Codex、Gemini CLI 等工具时,无需重新解释项目背景和之前的失败尝试。它试图解决当前 AI 编程工具“记性差”和“换工具即失忆”的痛点。

MCPScan

MCPScan

MCPScan 是一款出现在 Product Hunt 上的 MCP(模型上下文协议)生态工具,主攻 MCP 服务的安全扫描与治理。它出现的时间和赛道选择,比产品本身的早期功能更值得关注:MCP 正在从“能用”进入“敢用”阶段。

The argument that “AI structurally centralizes power” because it’s currently compute hungry ignores 125 years of super exponential growth in compute price-performance. Improvement in algorithms and continued hardware…

The argument that “AI structurally centralizes power” because it’s currently compute hungry ignores 125 years of super exponential growth in compute price-performance. Improvement in algorithms and continued hardware...

Replit CEO Amjad Masad 公开反驳“AI 必然导致权力集中”的说法,认为算力成本 125 年来持续指数级下降,现有缩放定律只是经验规律而非物理定律,未来 AGI 级能力未必永远依赖数据中心。这场与 Anthropic CEO Dario Amodei 的隔空对话,触及了 AI 技术路线和治…

Watermarking without quality loss is a bit unintuitive, doesn’t feel like it should work. I made this artifact with Claude to help me understand how it works, sharing in case it’s helpful. https://t.co/ALReoOH6tB http…

Watermarking without quality loss is a bit unintuitive, doesn't feel like it should work. I made this artifact with Claude to help me understand how it works, sharing in case it's helpful. https://t.co/ALReoOH6tB http...

Anthropic为满足欧盟AI法案要求,正在推进AI生成内容的水印方案;开发者Thariq用Claude制作了一个交互式工具,直观解释“无质量损失水印”这个看似反直觉的技术原理,并引发社区广泛讨论。

I 100% agree with Dario that using AI to cure diseases (and speeding up the regulatory approval of AI breakthroughs in healthcare) could bring 10x the benefit to humanity as everything else combined. https://t.co/1CkV…

I 100% agree with Dario that using AI to cure diseases (and speeding up the regulatory approval of AI breakthroughs in healthcare) could bring 10x the benefit to humanity as everything else combined. https://t.co/1CkV...

Anthropic CEO Dario Amodei回应外界关于他“渲染AI风险过多”的质疑,强调自己的公开表态是风险与收益并重;前Robinhood产品负责人Peter Yang则更进一步,称AI用于治愈疾病并加速医疗监管审批,其潜在收益可能相当于所有其他领域加起来的10倍。这场讨论的核心,是AI行业叙事正…