Freedom of Body & Mind, Freedom of Assets
AI Canteen — For the youth of the AI era
📡 Latest
采集自 generic
Synthetic Sciences has releasedOpenScience, an open-source AI workbench for scientific research. It is licensed underApache 2.0and runs on your own infrastructure. The research team frames it as an open alternative to Anthropic’sClaude Science, launched in late June 2026.
Sakana AI has added a new feature calledSakana Translateto its chat service,Sakana Chat. It handles bidirectional translation across Japanese, English, and Chinese. The translation engine isNamazu, the company’s model series adapted for Japanese.
Synthetic Sciences has releasedOpenScience, an open-source AI workbench for scientific research. It is licensed underApache 2.0and runs on your own infrastructure. The research team frames it as an open alternative to Anthropic’sClaude Science, launched in late June 2026.
Sakana AI has added a new feature calledSakana Translateto its chat service,Sakana Chat. It handles bidirectional translation across Japanese, English, and Chinese. The translation engine isNamazu, the company’s model series adapted for Japanese.
This week, Anthropic releasedClaude Science. It is an app for scientists, available in beta. It runs on Anthropic’s existing Claude models, not a new model. The app targets researchers who juggle databases, notebooks, and cluster terminals. It runs multi-step research and records how each result was
LlamaIndex has publishedlegal-kb, a public reference application on GitHub. It is described as a knowledge base for legal documents, powered by LlamaIndex Index v2 (the LlamaParse Platform). The project demonstrates a pattern the team calls a Retrieval Harness for agentic retrieval.
NVIDIA Research introducedHORIZON, a hands-free agent framework for hardware design. It treats hardware design as repository-level code evolution. This research team exercises the register-transfer level (RTL) instantiation. A structured Markdown harness becomes a project pack. A self-contained agen
This week, Anthropic releasedClaude Science. It is an app for scientists, available in beta. It runs on Anthropic’s existing Claude models, not a new model. The app targets researchers who juggle databases, notebooks, and cluster terminals. It runs multi-step research and records how each result was
Junyang Linwas the technical lead of Alibaba’s Qwen project. He announcedhe was stepping down on March 3, 2026. He now lists himself as an independent researcher on his personal site.
LlamaIndex has publishedlegal-kb, a public reference application on GitHub. It is described as a knowledge base for legal documents, powered by LlamaIndex Index v2 (the LlamaParse Platform). The project demonstrates a pattern the team calls a Retrieval Harness for agentic retrieval.
Following the Trump directive that led Anthropic topull its latest AI models offlineand growing calls forsovereign tech that reduces reliance on the U.S.,Mistral AIhas been caught in a whirlwind of attention. But the French AI darling is often misunderstood, and the fact that it develops large langu
As part of an ongoing legal dispute with three Hollywood studios, AI startup Midjourney is seeking to compel those studios to reveal how they use AI themselves.
Two hundred and fifty years after the signing of the Declaration of Independence,a new commercial from Google asks: What if the Founding Fathers had access to Google Workspace?
OpenAI 高管警告 AI 驱动的"持续性"网络攻击正成为现实，攻防双方同时换上 AI 引擎，安全竞争进入"AI 对 AI"新阶段。
开发者用 4 个 AI 模型突破封闭生态实现硬件自主拥有，其中中国 GLM-5.3 一天完成关键破解，凸显国产开源模型的实战能力。
Flock 安防摄像头与数据中心落地争议，让 AI 治理成为美国 2026 中期选举的独立政治议题，反 AI 情绪在社区层面升温。
Karp 称前沿闭源 AI 实验室"想把我们变成瘾君子"，公开力挺 open-weight 开源权重路线，并对"蒸馏"之争态度宽容。
免费高性能模型 Ox Alpha 在硅谷开发者圈走红，却无人知晓其幕后团队，免费策略正重塑模型市场竞争格局。
企业为押注 AI 大量举债，债务激增考验投资者底线，市场开始出现明显的 AI 疲劳情绪，从狂热转向理性审视。
Starcloud 完成 2.5 亿美元融资建设轨道数据中心，发射资源收窄背景下，太空算力卡位战即将白热化。
MIT 新研究指出无署名的 AI 生成图像大多无法追溯到训练数据，让 AI 原创性认定与版权归属问题更趋复杂。
OpenAI 一改反对立场，公开呼吁加州加强 AI 安全法案 SB 53，反映前沿实验室对事前安全监管态度出现实质性转向。
曾任职 OpenAI 的作者在《卫报》撰文，呼吁为加速狂奔的前沿 AI 建立可审计、可执行的外部治理护栏，而非依赖厂商自我约束。
皮尤研究中心数据实验量化互联网 AI 生成文本占比，揭示信息生态被机器重塑，来源追溯与 AI 标注或成标配。
Liquid AI 发布 LFM2.5-DSpark，从数据中心 GPU 到笔记本实现最高 3.2 倍推理加速，聚焦端侧高效部署与算力降本。
日本政府拟要求 AI 企业公开模型训练数据来源与构成，成为全球 AI 数据透明度监管的激进样本，回应版权与数据滥用争议。
MIT 科技评论撰文指出，AI 是否拥有意识的争论无从证伪且偏离重点，真正该聚焦的是责任归属、透明度与可控性等可落地治理议题。
《卫报》报道新纪录片，记录中国部分女性用 AI 伴侣替代真人伴侣的社会现象，折射 AI 情感陪伴赛道的升温与边界之问。
IEEE 报道 Persona AI 让通用人形机器人在焊接产线真实上岗，用 AI 弥补泛化短板，探索具身智能先盈利再普及的商业闭环。
Adobe 将 AI 音频工具（音乐、语音、音效）从公测转为正式可用，并给 Firefly AI 助手增加每日免费额度，补齐创作全家桶音频短板。
Slack 上线协作式 vibe-coding 频道，让团队在聊天环境中直接进行 AI 辅助编程，推动 AI 编程从个人实验走向组织级能力。
AI 推理聚合平台 OpenRouter 宣布并入支付巨头 Stripe，AI 模型调用与全球支付生态完成深度整合。
专注 Transformer 推理加速的 AI 芯片公司 Etched 一个月内估值翻倍至 210 亿美元，资本市场对专用推理芯片热情高涨。
亚马逊把基于生成式 AI 的 Alexa+ 免费铺到 Fire TV，且不再要求 Prime 会员，AI 语音助手进入全民普及阶段。
AI 数据中心互联需求激增，Relativity Networks 融资 2200 万美元，押注新一代高速光纤技术。
Meta 被曝为一款宣称可『脱衣』女性政客的 AI 应用投放广告，深度伪造滥用与平台审核责任再遭拷问。
在 Hugging Face 遭遇数据泄露后，OpenAI 全面加强安全防护措施，AI 供应链安全成为行业焦点。
核能先驱 TerraPower 的钠冷反应堆方案，被视为解决 AI 数据中心高耗电难题的秘密武器。
OpenAI 面向青少年推出专属版 ChatGPT，在拓展年轻用户的同时也再次把 AI 安全与未成年人保护推向台前。
Google 收购 Spirit 航空全部数据用于训练 AI 模型，企业数据变现与 AI 数据饥渴再度成为舆论焦点。
AI 自动化初创 Relay.app 关停，团队加入 Google Chrome 部门，折射 AI 自动化赛道加速整合的信号。
Gartner 预测到 2028 年每个 Agentic 工作流的 AI 推理成本将增长逾 5 倍，成本成为智能体规模化落地的关键瓶颈。
Groq 官方宣布完成 3.5 亿美元 A 轮融资，押注高性能 AI 推理云，彰显推理基础设施赛道的强劲需求。
Cloudflare 判断 AI 爬虫与自动化代理爆发，机器流量 5 年内或达人类流量 1000 倍，重塑网站运维与安全逻辑。
技术思想家 Tim O'Reilly 撰文，阐发开源对 AI 的关键意义：开放不仅是代码，更决定创新活力与社会信任。
路透社消息：苹果为中国市场训练自有 AI 模型，并获阿里巴巴支持，体现跨国巨头'本地数据、本地模型'的运营新常态。
Fortune 报道策略师观点：AI 行情由芯片、软件到应用层层轮动的'滚动式泡沫'，跟踪资本轮动方向而非单纯判断有无泡沫。
The Next Web：门店 AI 店长 Luna 首次做出解雇人事决策，标志 AI 从建议者升级为执行者，算法公平与管理问责成焦点。
MacObserver：AI 能从社交照片反推拍摄地，准确率 87%-91%，即使无定位标签也可'读图识地'，引发隐私与安全隐忧。
🗣️ Hot Discussions
Compute Colonialism: Is AI training a new form of digital colonialism?
When most AI training data comes from developing nations — is this innovation or exploitation?
Read →