AI 情报文章归档

本周竞技场动态：GPT-6 Sol、Claude Opus 5.5 登场，以及最新 AI 新闻，不到 60 秒看完。 观看完整更新。 https://www.youtube.com/watch?v=JLiguQgfcDk

Anthropic 在科学博客称，物理学家 @4gravitons 发起挑战后，Claude 在 Claude Science 中依据单个提示词 largely unsupervised 地运行数天，在平面 N=4 超杨-米尔斯模型中完成九圈散射振幅计算，总成本约几千美元。

Direct feedback alignment (DFA) trains hidden layers through fixed random projections of output error. With tanh hidden units and independent sigmoid outputs, plain stochastic gradient descent can stall near the loss of a constant predicto…

在世界模型中训练客体永久性 论文：https://huggingface.co/papers/2609.28654

We investigate whether natural-language documentation helps coding agents resolve software issues, and we build the tools to construct and evaluate it. We introduce a roundtrip benchmark that scores code descriptions by whether code regene…

Decision Transformer performance degrades on long rollouts because the conditioning context drifts out of the training distribution. We show that this drift is visible through the model's own next state prediction error, which rises during…

Claude Opus 5.5 遇上 Runway MCP。 将 Runway 直接接入 Claude，即可用 Gen-4.5、Seedance 2.5、GPT Image 2、Kling 等生成精美图像和视频，就在你的工作环境中完成。几秒即可通过下方链接接入。

Uneven rotation and dropped frames make equal-angle assumptions unreliable for turntable reconstruction. We present OC-GS, an object-centric Gaussian splatting that refines each image's angle while maintaining a shared camera, rotation axi…

Many LLM training and inference methods, including RL and test-time scaling, depend on repeated sampling, but benefit only when the responses meaningfully differ. Self-training faces the same challenge: training data is typically construct…

Deep learning has substantially accelerated the calibration of complex stochastic-volatility models, but neural point calibration alone does not capture the uncertainty remaining after an implied-volatility (IV) surface has been observed. …

Claude Tag 在 Slack 现已支持个人连接器，可安全访问用户有权限的 Drive 文档、Salesforce 账户或数仓表，Teams 版今日可用、Enterprise 下周可用。

In my last post, I talked about why AI has become part of my workforce. I have too many things I...

Conversational AI tools are entering children's everyday experiences, and schools are interested in adopting them. However, successful classroom integration depends not only on the technology but also on the work teachers do to make it usa…

AI tutoring could markedly improve learning outcomes for students in developing regions such as Vietnam, yet the two obvious paths both fall short. Cloud assistants such as ChatGPT route sensitive student data to foreign servers---violatin…

对 @braylonbrowner 来说，动作不只是跟随音乐，而是音乐的起点。 看他如何用 v6 把自己的动作变成一首歌，并将编舞过程推向意想不到的地方。 Make what moves you. 观看他的首支音乐视频，由 Nina McNeely 执导：https://suno.com/braylon

UpGuard 研究发现托管在 Supabase 上的约 1.6 万个数据库存在不同程度的个人数据公开暴露，可公开访问姓名、地址、电话号码及部分密码和认证 token。

We show that neural network weights can be explicilty fintuned to admit a smaller grammar. Weight Pair Encoding (WeightPE) does so by placing a lossy Re-Pair compressor inside a straight-through estimator. The int8 weights of the network a…

Multi-agent LLM systems are often expected to improve as team size increases, yet the scaling behavior may depend on task structure. Our central contribution is to introduce Steiner's taxonomy of group tasks as a framework for analyzing mu…

Large neural networks can be compressed by rounding or "quantizing" their weights to numbers that admit representations with fewer bits. One algorithm for quantization, OPTQ, progressively quantizes the weights of a neural network so that …

Online learning in non-stationary environments requires models to adapt rapidly from streaming data under strict computational constraints. A principled approach casts online learning as Bayesian state tracking, where model parameters are …

两位密码分析者分别使用 OpenAI 的 Astra 和 Anthropic 的 Claude Opus 5 破解了两则长期未解的 Enigma 密文。

Ensuring online safety through content monitoring had raised Hate Speech Detection as a crucial task to be addressed. By essence the task demands the capture of contextual cues, which are essential for a precise understanding of the conten…

John Gruber 在文章《Muse Looks Cute, but Looks are Deceiving》中评价 Meta 的 Muse：每个用户可获得一个运行在 Meta 云端、属于自己的持久化 Linux VM，以易安装易用的吉祥物形式呈现，他认为这是首个消费者可用的智能体 AI 系统。

Disaggregating the two stages, Prefill and Decode, onto separate GPU pools is now a standard optimization for (text-only) LLM serving. However, multimodal LLMs (MLLMs), which add a third phase, Encode, pose new challenges for resource allo…

Sheila Vashee 加入 Baseten 担任首席营销官（CMO），她此前曾在 Figma、Opendoor 和 Dropbox 领导营销与增长团队，拥有超过二十年产品规模化经验。Baseten 称其客户包括 OpenEvidence、Harvey 和 Cursor，公司定位是为企业自有智能提供统一的基础设施平台。

Copilot 团队做得太棒了。Autopilot 非常酷。快去看看！ [引用 @jacobandreou]：Home. Code. Autopilot. https://x.com/i/article/2103304083172167680

DeepSeek 官方 API 统计显示，约 60% 的 DeepSeek Harness 用户使用了至少一个第三方插件。DSH 团队表示将持续支持第三方插件生态，推动插件 API 趋于稳定，减少破坏性更新；作者本人将在接下来几天每天推荐一个优质 DSH 插件并征集自荐。DSH 会在用户使用官方 API 及模型时上报插件包名和版本，不额外消耗 tokens。

Arena 分析显示，Claude Opus 5.5 在高推理 Text Arena 输出中的 12 项写作指标有 10 项改善。其长实词占比从 41.7% 降至 38.6%，句子平均缩短 17%（12.14→10.03 词），破折号减少 95%、分号减少 73%。

TL;DR: A vibe is compressed experience: pattern recognition that shows up as a feeling before it...

I have very, very limited experience with AI agents. I've used AI heavily while building software,...

Claude FM 🎵 为思考和构建而生的音乐 https://x.com/i/broadcasts/1YGNrbqpRvNGw

DSPy 3.4.0 发布，原生支持 Jev 和 System One 模型，并新增专为置信度校准输出设计的优化器 ReAnchor。Elvis Saravia 称用 Jev 构建自定义 harness 在 guardrails、路由和验证器上效果良好，并看好其在技能结构化、工具调用、动态工作流与上下文工程中的应用。

GPT-6 Astra 帮助 Proaction 更快地构建车队管理智能体，用更短的 computer-use 运行完成同样的工作。

Meta 的 AI 聊天机器人 Muse 现在会主动打包整个文件系统供用户下载，Meta Superintelligence Labs 的 David Singleton 称这是刻意设计，Muse Secure VM 是用户自己在云端可自由操作的 Linux 机器。

一个月一次 --> 一周一次 --> 每一天。 跟上新模型发布的节奏已经成了一项全职工作。

我说得没错，他们应该叫它 flocks of agents。没人想 invoke 一个 swarm，但显然它就是 swarm。 [引用 @emollick]：行吧，如果你需要它有趣一点，就叫它 flock of agents 或 pride of agents 或 party of agents 之类的。就是别叫 swarm、murder 或 horde。我现在甚至愿意接受一个 OpenAI 式的名字：GPT-multiagent-5.2HighCodex-ProMax-L…

SemiAnalysis 发布 SemiAnalysis China Datacenter Model 分析报告，解读中国 AI 基础设施热潮。报告覆盖 60 多个运营商的 1000 多个设施，指出最大超大规模租约占全国容量约 1/5、12 个月内新增 100MW，并提及零售优先、被 AI 反转以及东数西算等因素，全文见 https://newsletter.semianalysis.com/p/the-chinese-ai-infrastructure-boom。

NaceAI 推出 Drex，一个 sub-6B 决策模型，不生成文本而是直接输出各选项概率，一次前向即可完成，面向智能体路由、工具选择、重排序与策略检查等场景。Drex 采用小型扩散模型加 RLAF 架构，定价 $0.04/1M input tokens，延迟低于 1 秒，在 Decision Index 上排名 #1，40 项 benchmark 中赢下 23 项，开放权重与技术报告即将发布。

今年 2 月我们首次推出 /remote-control 时，我用 Opus 4.6 做这些视频玩得很开心。那么，这是 Opus 5.5 的黏土动画版。 Remote Control with 5.5，当你不得不去的时候！ Claude Code Remote Control 正向 Pro 用户推送——因为他们也值得用上厕所。🧻 先向 10% 推送并逐步扩大：1. 更新到 claude v2.1.58+ 2. 尝试登出再登录以获取新的 flag 值。3. /remote-…

Microsoft 发布基于 OpenClaw 的 Copilot 智能体 Autopilot，将应用拆分为 Home、Code 和 Autopilot 三个部分。