#01
Claude Code's "Don't Ask Again" exposes passwords in 1,122 public GitHub files
推荐理由该成果可用于警示并排查因在 Claude Code 等智能编程工具中跳过授权确认而造成的代码库泄露隐患，全面清理公开发布的生产环境 Azure 数据库密码、云管理员密钥、API 令牌及 SSH 登录权限，防范核心基础设施被攻击者直接接管。
Pai 的 AI 雷达
含近几日仍热话题，按时间衰减排序。今天先看 HelpCode-ai/anythingmcp，其次 chiphoton/DeepSeek-Harness-Video-Director。并行还有 Claude Code's "Don't Ask Again" exposes passwords in 1,122 public GitHub files、Copilot code review: API management and review overview updates。
You can now manage GitHub Copilot code review settings through the REST and GraphQL APIs, from enterprise policy down to individual repositories. Copilot’s review overview comment is now more compact, and this update includes fixes for billing, automatic reviews, and reliability.
The Agent adoption chart in the Copilot impact dashboard now includes Copilot Chat and the VS Code Agents window, so you can see engagement across nine Copilot experiences. The adoption cohort called “Passive users” is now “Chat-first users”, which better describes who is in it.
📌 一句话摘要 本文从工程视角系统阐述了构建稳定 AI 人格系统的方法论，提出三层 Persona 架构（角色基线、人格系统、风格系统），并提供可复用的模板和验证方法，帮助开发者打造可控、稳定的 AI 智能体。 📝 详细摘要 文章聚焦 AI 智能体工程中的核心问题——人格（Personality）与风格系统（Style System）的设计。首先指出 LLM 天生具有可塑性，若不加约束会导致输出漂移、风格不稳定。随后从工程角度提出三层 Persona 架构：Layer 1 为角色基线（定义角色、核心能力、责任范围、禁止事项），Layer 2 为人格系统（
When deploying diffusion models in design tools, image editors, and visual production workflows, developers must account for inference latency. Generating an image requires multiple sequential passes through a model with billions of parameters, and higher resolutions increase att
The "Into the Unknown" benchmark tasks nine AI agents with tackling open research questions from U.S. national laboratories. None of the 135 submissions meets all requirements. A total of 114 papers fail the integrity check, with agents claiming to have conducted experiments that
During reinforcement learning (RL) post-training, training trajectories are generated on the fly by running inference with the current policy model. For training efficiency, it is common to run inference and training in parallel on separate GPU resources in a disaggregated setup.
Profession like software engineer, mechanical / graphics designer takes a lot of strain on your eyes, observing minute details for hours and hours on screen can significantly reduce the blink rate which you might not even notice until it gets irreversibly painful, leaving you wit
A little web game I made as an anniversary present for my wife. Upload vacation photos and make them into a game. WWW uses EXIF to determine where and when the photo was taken. This is then turned into a game you can send to anyone.
AI models can often spot problematic behavior by other AI agents in logs, according to a study by Anthropic researcher Andy Hall. GPT-5 found about 98 percent of reward-hacking cases, but the more exhaustively the search is tuned, the more false alarms pile up. That's not yet eno
📌 一句话摘要 文章聚焦三个哈佛女孩（郭文景、Christine Zhang、Avery Pan）在 AI 创业领域的崛起，分别创立了 Pika、Veil 和 Class Companion，展示了新一代华裔女性如何把握 AI 浪潮实现个人价值与商业突破。 📝 详细摘要 文章以二十多年前的哈佛女孩刘亦婷为对照，引入三位新生代哈佛女孩在 AI 赛道的创业故事。郭文景创立 Pika Labs 做 AI 视频生成，完成 1.35 亿美元融资，带动信雅达股价上涨；Christine Zhang 创办 Veil 做生成式引擎优化（GEO），100 万美元种子轮快速
For Claude to learn from experience, it has to decide what’s worth remembering before it knows what’s next. Here’s how “dreaming” helps it organize memory and correct mistakes.
OpenAI is partnering with Hacktron to find vulnerabilities, test connections with third-party services, and strengthen defenses with frontier AI.
Analysis of over 913,000 coding agent sessions reveals that 13% expose agents to inline TODO(b/...) comments. In 6% of encounters, agents act on these comments, with 83% being entirely unprompted by the developer. These detours result in significant overhead—adding 70% more steps
Open-source AI coding agent Discussion | Link
One workspace for your team and its AI coding agents Discussion | Link
It’s a relatively quiet news day, and the AI market rout we referenced obliquely two days ago has deepened, though it might be leveling off not even halfway into bear market territory (-10% on Nasdaq, not -20%). That said, everything being “priced to perfection” means any surpris
周末也值得一看的新闻 触屏 MacBook Pro 开卖首日，部分渠道缺货，部分渠道发券 苹果向部分 iPhone Duo 买家发问卷，询问折痕和屏下相机体验 仰望 U9 Xtreme 在阿布扎比跑出 503.17 km/h 新华社与优必选发布仿生机器人主播，张艺谋负责形象设计 张雪：将参加 36 名重庆大学学生的校招二面 马斯克称特斯拉得州工厂将缩减 Cybertruck 生产，为 Cybercab 腾出产能 小米汽车国际业务部架构曝光，销售、交付和售后分设团队 广电总局整治微短剧「盗脸」「偷声」，换名重播也在治理范围内 中国空间站把铷原子云冷却到 9
You might already run Claude Code locally on your laptop, backed by Amazon Bedrock. This post shows how to go a step further: run it as a remote agent on Amazon Bedrock AgentCore, a platform to build, connect, and optimize agents at scale, with any framework or model. The sample
Fabian Theis has been one of the key players in the various iterations of the Virtual Cell concept and initiatives. It is not hyperbole at all to say that his lab has been pushing the frontier of algorithms to understand multi-scale, multi-modal (”multiomics”), time-varying biolo
Concerns are being raised that neither the Commerce Department nor Nvidia are doing their due diligence when it comes to vetting Chinese firms cleared to handle the company's H200 AI chips. According to The New York Times , New Matrix is on the list of approved firms, yet its off
上午跟吴熳老师录了一期播客，播放地址 https://www.xiaoyuzhoufm.com/episode/6aed7a995b7c4709bfe1fdae Claude Opus 4.6 基于音频转写生成一篇文章。 树村，北京西二旗附近一个没有水泥地的村子。一下雨，全是泥泞，根本没法走。 但那里住满了搞摇滚的人。 打鼓的、弹贝斯的、弹吉他的，谁都别嫌谁吵，村民也习惯了。 房租一个月 200 块，还能拖欠。 痛仰这样的知名乐队，在酒吧演出一场，整个乐队分 500 块钱，五个人，一人 100。 演出结束舍不得打车，拿一瓶大绿棒子跟粉丝喝到天亮，第二天坐
Webinar Recap: From Data to Vectors to Context
PLUS: Apple’s first robot faces another delay
当我们想喝可乐的时候，在极长的时间都只有两个选择：百事可乐和可口可乐。当我们选择手机的时候，有 90%的概率在苹果和华米 OV 等品牌里辗转。当我们买运动服饰的时候，第一时间想到的，大概率是 Nike、Adidas。 但世界之所以缤纷多彩，是因为在这些巨头之外，有一些不遵从传统，力求创造不同，注重设计和功能，着眼于明日的公司存在。 它们有着非主流的商业模式，设计、技术与产品能提供独特的用户价值，和足够的社交谈资。重点是，它们没有大公司的包袱，敢于不顾一切的进步。它们，是「多样性公司」。 多样性，是开放世界的关键。爱范儿相信，只有真正关注和理解多样性公司才