全部 AI 动态

Kling 4.0 Makes Its Debut on OiiOii. Generate Videos up to 30s, Cinematic Quality, Precise Motion & Stable Scenes. This is Kling 4.0. Try now and get 25% OFF until Nov. 30.

智能体会写代码了，工程师还剩什么？ 吴恩达老师基于 1w+ 职位分析与数十次专家访谈，用五封公开信绘出「AI 工程技能地图」，答案分四层： 构建与部署不可预测的 AI 应用、守住软件工程根基、高效驾驭编程智能体，最终上移到塑造构建：决定该做什么，并对结果负责。 https://x.com/i/article/2098560800181125120

Kling 4.0 现已在 OnSolo 上线。 一次生成 30 秒原生视频。 在快速动作和大范围镜头运动中保持稳定。 立体声，唇同步更精准。 21:9 宽屏。单个任务最多可叠加 15 个参考和 10 个关键帧，让你的角色、镜头和故事保持一致。 今天就用它制作你的下一部短剧、电影或 FMV。

使用 GPT-5.6 Sol 的研究人员在 PHP 中发现了一个极其严重的漏洞，PHP 驱动着约 70% 的网站，该漏洞导致主要网站下线，运营方纷纷紧急修补。 发现该漏洞的 AI 安全公司 Calif 在一份声明中表示，如果不尽快修补，这些漏洞可能使全球 70% 的网站面临风险，并称其为“AI 发现的最严重漏洞，甚至可能是整个软件史上最严重的漏洞”。 > PHP 紧急发布了修复程序，因为这些漏洞可能让无账号的攻击者接管 Web 服务器。 > 两个漏洞在 CVSS 严重性评分中均为 9.2 分，影响低至 5.4 的 PHP 版本。 > 当请求在错误的时机触及 PHP 的内存上限时，PHP 会释放内存但保留指向该内存的指针，该工作进程上后续的请求便可劫持这块内存。 > 审计人员在 2024 年发现了这一弱点，但修复工作停滞了数月，部分原因是没有人展示过实际攻击。 > 现由 Google 拥有的云安全公司 Wiz 表示，这些模型让一名研究人员在一周内完成了过去顶尖团队需要数月才能完成的工作。 > CVE 编号的泄露迫使 PHP 提前一天发布，Cloudflare 目前报告称已出现早期攻击尝试。

Set your wildest ideas into motion at the speed of thought. Made for every screen, shaped for every beat. Kling 4.0 is now available in Firefly. Try it today: https://adobe.ly/4kf6afg

Grok Bot has saved me at least 12 hours of social media reporting this week. Especially now that it has access to X and can do sentiment reporting, pull real time analytics, and tell me about trends on specific occasions. Combined with the ability to plug into @NotionHQ and have my team informed while I’m literally sleeping is crazy. Vs the past where I would’ve had to set alarms every few hours or wake up at 4 am to prepare an update for the company. Well done @larsencc and team Such a great experience.

CS 329Z: Engineering AI Agents Stanford / Fall 2026 @stanfordnlp https://cs329z.stanford.edu/ 课程定位：从"模型"到"系统"的工程学 覆盖：简单 LLM 流水线 → 复合 AI 系统 → 自主 Agent。三位讲师的背景也高度互补： @Diyi_Yang（斯坦福 NLP 教授，人机交互与社会计算方向） @michaelryan207（DSPy 核心贡献者，自动评估 AutoMetrics 作者） @jyangballin（SWE-agent / SWE-bench / SWE-smith 作者，软件工程 Agent 领域最重要的研究者之一） # 课程主线：三大工程挑战 贯穿全课的三个核心问题——分解（decomposition）、数据（data）、评估（evaluation）。11 周的内容基本围绕这三条线展开，可以分为五个模块： 模块 1：构建基元（Week 2–3） · LLM 作为构建材料：API/SDK（litellm）、结构化输出、约束生成、解码策略、test-time compute、上下文工程、模型选型与成本/延迟权衡 · RAG：embedding、向量库、分块策略、混合检索、cross-encoder 与 ColBERT 后期交互 · 工具调用：函数调用 API、MCP（Model Context Protocol）、工具设计、代码沙箱、错误处理与重试 模块 2：框架与设计模式（Week 3–5） · 框架层：DSPy（signature / module / optimizer）、LangChain/LangGraph、LlamaIndex，重点是"框架抽象了什么 vs. 你手写了什么" · 设计模式：workflow vs. agent 的分类学，五种可组合 workflow 模式，ReAct / plan-and-execute / reflection，"scaffold（脚手架）本身就是设计决策" · 记忆架构：短期/长期记忆、记忆作为工具动作、文件系统作为外化记忆、跨 Agent 记忆（MemGPT、Mem0、Generative Agents） · 多 Agent 系统：编排模式、handoff 与状态传递，以及一个很有态度的对照阅读——既读 AutoGen，也读《Why Do Multi-Agent LLM Systems Fail?》和 Neubig 的《Don't Sleep on Single-agent Systems》 模块 3：优化（Week 5） · 从提示词到微调的全景：GEPA、MIPROv2、OPRO、TextGrad（提示优化）；LoRA/QLoRA、蒸馏、RLHF/DPO（权重优化）；test-time scaling（推理算力） · 核心问题是决策框架：什么时候优化 prompt、什么时候优化 weights、什么时候堆推理算力 模块 4：数据与评估（Week 6–8）——最有分量的部分 · 数据：trace、demonstration、feedback 三类数据；训练数据 vs. 评估数据；数据飞轮；合成数据；从 Agent 轨迹构建数据集（SWE-smith） · 评估基础：为什么 eval 难；4 元组框架（request / environment / stopping criteria / scorer）；好 benchmark 的性质；tinyBenchmarks · 评估基础设施：三类 grader、LLM-as-judge 的 prompt 设计与已知偏差、pairwise vs. pointwise、非确定性指标 pass@k vs. pass^k、harness 设计 模块 5：安全与前沿（Week 8–11） · 安全：工具访问的隐私风险、prompt injection（含间接注入）、红队、沙箱与权限模型、输出护栏、human-in-the-loop · Coding Agent：SWE-agent、Claude Code、OpenHands 的端到端架构对比 · 主动式 Agent：从 reactive 到 proactive，General User Models（GUM）、Next Action Prediction，以及"Agent 何时应主动、何时应等待"的 mixed-initiative 问题 · 开放问题：多模态/web/计算机使用 Agent、科学 Agent、长时运行架构、生产可观测性（tracing、monitoring、成本管理） # 作业设计：一手建、一手评 HW1：从零构建 Agent 系统（10%） 给定论文库，构建能检索并推理回答科学问题的 Agent。 · Part A：只用 litellm 手写 RAG + 工具调用 + ReAct 式 Agent 循环 · Part B：用 DSPy 重建关键组件，并反思框架抽象了什么 HW2：评估一个 Agent（10%） 给定一个预构建 Agent，设计完整评估套件：代码型 grader、至少一个 LLM-as-judge、用 4 元组框架构建 benchmark 任务、错误分析。

There's not enough buzz around Claude Mods. Read this Claude Mods article if you want to: - Understand Claude Mods basics - Install and use the most useful mods today - Ponder where the agent-human interface will go For @latentspacepod.

When our AI code reviewer's cost per run doubled, we dug into the token usage to find out why. Learn how we used prompt caching and intelligent model routing to slash our inference bill by 68% while speeding up reviews. https://cfl.re/4isj2NU

Introducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine. • Best-in-class fast local coding • Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark • Built from scratch in under 4 months Private, unmetered intelligence on your own machine. Open weights available now on @huggingface, run it locally on your favorite framework or try it for free on @opencode or @pidotdev.

I just published my roadmap for building evals on highly subjective tasks. Here are the key steps involved and practical tricks for reducing noise / increasing confidence… Blog is here: https://cameronrwolfe.substack.com/p/subjective-evals Why is this difficult? Evaluating agents is already hard due to their broad scope, emphasis on expensive / long-horizon tasks, and non-determinism. Subjective tasks make this even harder because we cannot deterministically verify solutions and must rely on human judgments (or LLM judges). This introduces additional noise into an already difficult evaluation process. Evaluation roadmap. We can build reliable evaluations using a repeatable framework with three components: - Human evaluation: Our North Star for quality. Create quality guidelines that clearly define quality with the help of subject-matter experts, calibrate human evaluators to ensure agreement, and continually monitor evaluation quality. - Offline evaluation: Use calibrated LLM judges or task-specific rubrics as an efficient proxy for human evaluation. This allows us to scale experimentation without relying on expensive human evaluation for every new model. - Production monitoring: Track quality signals from real-world agent usage and use A/B testing to validate improvements before fully deploying a new agent. Multi-stage evaluation. These components form a tiered evaluation process: Offline evaluation → Human evaluation → Production A/B test → Full deployment We can test many ideas automatically, send promising candidates to human evaluation, and validate improvements through production A/B testing. Continual production monitoring then allows us to discover new use cases / failure cases in production that we can add into the earlier stages of evaluation to create a feedback loop. Sampling useful data. Evaluation quality depends heavily on which tasks we select. We want tasks that are diverse, realistic, difficult, discriminative, and accurate. Techniques like structured task taxonomies, model-based difficulty filtering, Item Response Theory (IRT), and LLM-based auditing help us construct more informative evaluation sets. Handling noise. We should leverage statistics throughout the evaluation process: - Power analysis: Estimate the number of evaluation samples needed to detect meaningful improvements. - Confidence intervals: Quantify uncertainty and provide clear confidence signals alongside evaluation results. - Re-sampling: Repeat agent / judge evaluations to reduce within-task variance and potentially tighten confidence intervals. - Paired comparisons: Evaluate models on the same tasks to more reliably detect performance differences. Key takeaway. An improved benchmark score does not always mean our agent is more capable. We need well-designed evaluation tasks, calibrated human / model evaluators, and statistical confidence signals to distinguish meaningful capability improvements from noise (especially on subjective tasks). By combining these techniques into a repeatable evaluation process / roadmap, we can make building and maintaining evaluation suites more systematic, reliable, and scalable.