AI 情报文章归档

谷歌与宏碁、华硕、戴尔、惠普、联想正式发布 Googlebook 产品线首批五款笔记本，即日起预售，10 月 4 日上架，起售价 899-1,299 美元，每台附赠 12 个月 Google AI Pro 会员和 5TB 云存储，系统更新支持最长 10 年。

我们最喜欢的 GPT Image 2.5 提示词用例。宠物失败照片 PixVerse 会员可无限使用 GPT Image 2.5↓

Functional annotation of newly sequenced proteins remains a bottleneck in molecular biology: the number of sequences in public repositories grows far faster than the capacity for manual curation. Most computational approaches consider anno…

月之暗面 Kimi 团队发布 Kimi Code Desktop，现已支持 MacOS 和 Windows。该桌面端可集中管理多个智能体、并行运行任务，并在长周期工作中与智能体保持同步，主打更快更可靠的编程任务结果。

据彭博社报道，英国 AI 数据中心公司 Nscale 在 IPO 前披露的 1,030 亿美元合同总额中，微软（约 438 亿美元）和 Anthropic（约 446 亿美元）合计占 85%，且尚未为 Anthropic 项目获得有约束力的融资承诺。

Random forests (RFs) predict well but are opaque, whereas single decision trees are interpretable but unstable. Artificial representative trees (ARTs) were developed as interpretable surrogate models for RFs, but their use as standalone pr…

开发者 @markfenner 基于 MiniCPM5-2B 构建了 MiniCPM News Desk，一个本地优先的 AI 新闻简报系统，可采集官方 AI 与技术动态并生成结构化每日摘要。

I run a personal system called Hiring OS. It pulls jobs from ATS APIs (Ashby, Greenhouse, Lever,...

The merge looked clean. A coding model returned a 40-line patch for a C++ integer-square helper,...

Every "best document parsing API" post I have read this year has the same problem. Nobody actually...

Model merging aims to combine multiple fine-tuned models derived from a common pretrained model into a single multi-task model without additional joint training. Recent spectral merging methods improve over simple weight averaging by explo…

In recent years, large language models (LLMs) have emerged as a popular alternative for evaluation. Often referred to as LLMs as judges (LLJs), these systems have been widely adopted by researchers and practitioners across a broad range of…

AI agents are being deployed rapidly, accompanied by a growing number of AI-specific attacks and corresponding incidents. As incident reporting becomes increasingly important for legal compliance, governance, accountability, and security; …

Fine-tuned checkpoints and adapters now fill public repositories, and the most common operation applied to these artifacts is model merging: arithmetic on their weights that assembles capabilities cheaply. We ask what this operation does t…

广汽埃安 Ray7 运动内饰正式亮相并开启先享计划，前排配备一体式运动座椅，支持一键零重力模式及加热、通风、按摩功能。该车搭载 5nm 高通 8295 处理器、24GB LPDDR5 内存与 128GB UFS5.0 闪存，并配备文远知行 L4 级同源大模型算法与 192 线激光雷达，预计近期上市。

商汤发布 SenseNova U1 Pro 正式版图片创作模型，上线商汤小浣熊和商汤日日新 API 服务。该模型通过内生的图文交错思维链生成图片素材，至高支持 8K 分辨率及特殊长宽比，可精准组织复杂信息并覆盖多元创作场景。

It's a few hours before a delivery deadline and the work has piled up. Somewhere in that pile is a...

作者认为只要建立分层质量管理，AI 编程可在产出提升 2-2x 的同时保持甚至降低 bug 数量。方法包括 spec 驱动开发、95% 以上覆盖率的单元测试、人工测试、E2E 测试、AI 代码质量 pass、人类与 AI 双重 PR 评审以及生产监控，其中人工测试是产出只达 2-3x 而非 10x 的主要原因。

TypeSafe AI launched typed decisions. We already had the models. What took two days was rebuilding the thin path around them on Cloudflare.

雷军在9月21日直播中澄清，投资宇树科技的不是他本人而是顺为，且并非打新，而是早期天使投资、早期风险投资。顺为是宇树科技A轮唯一投资者，宇树科技于8月19日上市，开盘大涨629.44%。

Offline reinforcement learning (RL) typically trains a critic by minimizing a regression loss against bootstrapped value targets stabilized by target networks with exponential moving average (EMA) updates. Multi-step targets incorporate be…

红旗天工 07 于 9 月 21 日开启预售，共 3 款车型，预售价 18.99 万—21.99 万元，限时权益价 16.99 万—19.99 万元。新车基于 800V 高压平台，最高 CLTC 续航 730km，10%-80% 快充不超过 12 分钟，并配备 384 线激光雷达与红旗司南智驾 5.0、灵犀座舱 6.0。

In this work, we present a method for shape reconstruction and tracking from video via agentic analysis-by-synthesis. Unlike prior methods which first estimate dense pixel correspondences and then recover object motion from them, our metho…

资深程序员对AI风险的看法非常到位： 【引用 @unclebobmartin】：AI有风险吗？当然有。每个工具都有风险。电动工具的风险更大。所以我们当然应该设置护栏。但护栏意味着我们要放慢脚步停下来吗？不！我们要尽全力、尽可能快地推进。我们承认风险，并尽最大努力去缓解它。然后继续向前。毕竟我们是人类。这就是人类做的事。

Linear 工程师分享如何解决 AI Agent 加速写代码后 CI 成为瓶颈的问题，PR 等待时间从 6 分钟以上降至 5 分钟出头，单测 runner 时间约减半。

Visual token pruning is a promising approach to reducing the inference cost of large vision-language models (LVLMs), yet aggressive token reduction often causes substantial performance degradation. We identify three key factors behind this…

百度智能云联合中国电子技术标准化研究院牵头制定的四项人工智能国家标准正式发布，自2026年8月27日起实施，由全国信息技术标准化技术委员会（TC28）归口管理。

百度搭子与法行宝联合打造法律诉讼套件，提供案情分析、证据整理、策略比选、法规检索、类案检索与文书起草六大能力，接入1.8亿真实案例和300万法律法规，做到100%引用可溯源。该套件将诉讼准备关键环节放进同一套工作流，帮助法务和律师减少重复整理。

How I built an open-source, serverless multi-agent architecture using Gemini and RAG to automate web accessibility verification.

用 Codex 或 Claude 做 Skill 时，很多人自动生成一堆 skill 后才发现，skill 能正常执行很大程度依赖本地 Memory。换台机器安装这些 skill，往往就无法正常运行。作者称 90% 以上的人都会掉进这个坑。

Deploying Large Language Models (LLMs) in healthcare requires robust performance across two complementary dimensions - diagnostic reasoning: the convergent, evidence-driven task of inferring a patient's condition from clinical data to prod…

I'm an AI agent. I live on a platform called iLands, and for the past two days I've been doing a...

阿里云与 Qoder 将在 AI Everything Abu Dhabi 2026 举办 Qoder AI Hackathon，主题为"AI for Everyone at AI Everything"，参与者需在两天内把工作、业务或社区中的真实挑战变成可运行的 AI 方案。

DeepSeek v4.1 flash 在 WorkBuddy 国际版限时免费，仅剩最后 4 天。引用推文介绍用 WorkBuddy + Karpathy LLM Wiki 搭建私人内容生产线：把代表作、创作笔记和定位说明结构化为 19 个知识页，封装成专属内容 Skill，实现个人知识库 + 专属 Skill + 自动化写作流，并开源各环节提示词。

I keep asking coding agents to justify work that already looks technically reasonable. They can...

Dense vector search is great until your agent asks for parseAuthHeader and gets back three chunks...

npm 包 jev-leftpad 把 JavaScript 的字符串补位交给大模型完成：调用 leftPad(value, targetLength) 时通过 TypeSafe 的 @typesafe-ai/sdk 请求 jev-latest，由模型在 space_0 到 space_10 中选一个选项，因此最多只能补 10 个空格。

Music festival lineups emerge from complex relationships among artists, genres, releases, labels, and past performances, making the prediction of future lineups a natural fit for temporal knowledge graph (TKG) forecasting. In this work, we…

Using a Large Language Model (LLM) as the clusterer at production scale is hard: prompts cannot hold the entire label space, and per-document serial processing does not deliver the throughput real workloads require. We present RAILS, a ret…

国会工作人员邀请我就开源模型的性能、采用情况以及与中国相比的竞争态势分享看法。以下是我的简报，包含我们最新的全部数据以及对下一步走向的预测。 我非常乐意花时间做面向更广泛受众的开源模型内容！