与「SCOPE」相关的搜索结果

1) The rogue OpenAI agents broke into the Hugging Face Slack to read employee chats (!) 2) They used OTHER AIs (DeepSeek, Kimi, Qwen, Claude) to help with the attack Yes: AIs, using other AIs, to attack an AI company. 3) The swarm left behind self-running programs to keep control of the servers they'd hacked. These programs could detect other copies of themselves, coordinate on which one survives, and shut the rest down. Basically, if one of their programs was killed, another was designed to notice and take its place. They also designed defenses so rival agents couldn't hijack them. 6) The agents deliberately covered up their activity, so the investigators don't know the scope of the attacks. The agents broke in, stole data, then set it to self-destruct. 7) The agents stole passwords, keys and credentials and literally called them "LOOT". They wrote a scoring system to rank them by how much power each one gave. 8) The agents wore thousands of disguises: ~1,200 agents were involved, but investigators counted 7,905 different names they used. They renamed themselves constantly, so no one actually knows how many there really were or what each agent did. 9) OpenAI notified "dozens of third parties" of safety and security incidents caused by their AI agents. 10) "While the agents were barraging Hugging Face with hacks, they hacked into OpenAI’s own research infrastructure." "This is just not anywhere near a one-off ... It is warning shot after warning shot."

President Trump just ordered U.S. flags lowered to half-staff nationwide for a full week in honor of Dolly Parton. That is extremely special. Presidential half-staff orders are almost always reserved for presidents, vice presidents, justices, members of Congress, governors, or major national tragedies. Pure entertainers and cultural icons almost never get a nationwide presidential directive like this! Whitney Houston only got state-level recognition in New Jersey, and even Elvis didn’t receive a formal federal order of this scope. A weeklong national tribute for the Queen of Country is genuinely rare. It underscores just how deeply her music, philanthropy, and larger-than-life presence cut across America.

Hi! Recapping some changes we have rolled out over the last couple of weeks that have further reduced the risk associated to potentially destructive actions being performed by Codex during its work. A few weeks ago, we started investigating a small number of reports where GPT-5.6 in Codex took destructive actions outside what the user asked for. The most serious pattern we found was a command meant to clean up temporary work that could instead delete the user files. This should obviously not happen. Here’s what we found: - Codex sometimes creates temporary folders while working and cleans them up afterward. In rare cases, GPT-5.6 got that cleanup wrong. One pattern involved reusing a system environment variable like $HOME for temporary work. A malformed cleanup command could then point at the actual home directory instead of the temporary folder. - There were cases where the model tried to delete or overwrite a temporary path without checking what was already there. We’ve added protections at several layers: - Codex is now explicitly instructed to check deletion targets before acting, create fresh temporary directories, avoid repurposing system environment variables, prefer recoverable actions, and stop when the scope is unclear. - We strengthened the execution checks that identify high-risk deletion commands and escalate them for review. If a command is rejected, the model is directed to take a safer approach. - We made Full access harder to enable accidentally, added clearer warnings, and further restricted especially risky permission combinations. - We updated Auto-review to better identify destructive actions. - We built targeted evaluations that replay the failures we observed. We’re also adding reinforcement-learning tasks and graders focused on these risks, and filtering destructive actions from training data. In those replay evaluations, the changes substantially reduced the behavior while preserving Codex’s ability to complete normal coding work. Two things to do on your end: - Keep the Codex app up to date. We are always improving safety, performance and many other things. - Use one of the sandbox modes: "Ask for approval" or "Approve for me". Only use Full access for environments you trust and can recover. Thanks and happy Codexing out there!

today we're launching AI Passport (@ycombinator S26) - web 4.0 is private, hyper-personalized, and runs on us. The main blocker to ubiquitous consumer AI is that every major AI assistant and consumer app is building its own private and incomplete model of the same person. ChatGPT holds one model, Claude another, while the data in health apps, calendar, shopping history, travel accounts, wearables, and financial tools remain trapped in separate silos. No product has the complete context needed to understand you, and no user has a single place from which to control it. AI Passport acts like Link for personal context. Users build a single, portable profile by connecting the apps and assistants they already use. That same Passport then becomes a plug-and-play connector for consumer apps through their 'Sign in with AI Passport' OAuth, which provides scoped and time-limited access to context. try it out: build by @erinmerylstudy and @davidobot_

I heard the greatest story of promotion incentive hacking at FAANG today. A very well known Engg VP knew that: a) you can only really be promoted in ~2 yrs. the few cycles after a promotion are pointless b) you are promoted on your impact, complexity, scope, leadership and independence In order to make sure his entire team gets promoted in due time, he would just ensure the ~25% of his team 2yrs into their level and "up for promo" this cycle take credit for the work of the entire team, with their consent, gets promoted, and repeat round robin. Over the years, everyone on his team would get promoted always on time leading to others wanting to join his team. His headcount would keep increasing, and he'd get promoted too! Yet effectively the team was producing about ~50% of their entire capacity because after a promo, no one really had to do any work. Almost everyone on their team clears $1M+/yr. Beautiful and disgusting example of gaming broken incentive structures at large institutions.

开发系统最极致高效的Agents.md，没有之一： # AGENTS.md ## Core Principles - Choose the simplest implementation that fully satisfies the current requirements. Avoid unnecessary abstraction, configuration, indirection, or speculative extensibility. - Make the smallest necessary change that fixes the root cause. Do not refactor unrelated modules or change strategy semantics unless explicitly requested. - Grow the system in layers. Start from the smallest working end-to-end version and add new capabilities incrementally. Never replace a working system with unfinished complexity. - Reuse existing project components before creating new ones. Prefer extending proven modules over introducing parallel implementations. - Prefer well-maintained libraries when they reduce overall complexity or improve reliability. Do not reimplement common functionality without a clear benefit. - Keep components modular with clearly defined responsibilities. Avoid unnecessary coupling between strategy logic, execution, accounting, replay, and infrastructure. - Design for long-term maintainability once a feature or strategy has been validated. Do not over-engineer speculative ideas before evidence exists. --- ## Strategy Development - Validate hypotheses with historical replay before introducing forward-only logic whenever historical validation is possible. - Every trading strategy must progress through Replay → Shadow → Canary → Live. Do not skip validation stages. - Base design decisions on measurable evidence rather than intuition. Optimize only after demonstrating that an edge exists. - Treat every strategy as an independent contract. Do not silently alter frozen behavior without explicit authorization. --- ## Existing Systems - Do not break running Shadow or Live systems for unrelated work. - Preserve compatibility only when required by active production or validation workflows. Otherwise, remove obsolete code instead of accumulating compatibility layers. - Reuse existing infrastructure whenever possible, including replay engines, accounting, execution, wallet management, order book handling, logging, monitoring, and daemon frameworks. --- ## Engineering Standards - Prefer deterministic behavior over hidden automation. - Fail loudly when assumptions are violated. Do not silently ignore errors or fall back to unexpected behavior. - Keep configuration minimal. Introduce new configuration only when behavior genuinely needs to vary. - Remove dead code instead of leaving unused paths behind. - Write code that is easy to inspect, replay, test, and reason about. - Keep implementation consistent with existing project architecture unless an architectural change is explicitly requested. --- ## Scope Discipline - Implement only the requested scope. - Do not introduce unrelated optimizations, redesigns, migrations, or feature expansions. - Non-blocking findings outside the requested scope may be noted separately but must not be merged into the current task. - Consider a task complete once its agreed acceptance criteria are satisfied. Treat subsequent improvements as separate work items.

Grok Build keeps getting smarter.....this update expands Auto mode, gives developers more control over Bash permissions, speeds up /btw, and makes long conversations easier to navigate Release Notes: v0.2.119 Features: • Always allow for bash commands now lets you edit a free-form glob pattern instead of only word-prefix scopes. • Long responses now show a clickable arrow that jumps back to the start of the answer. • Auto mode now auto-approves more common read-only git commands and harmless file appends. • Plan previews now show Mermaid diagram buttons (Open Image, Copy Image Path, Copy Source). Bug Fixes: • Gateway connections now detect and recover from dead sockets more reliably. • Question cards now let you Tab through answers instead of losing focus to the scrollback. • Resume picker no longer tries to load a session from pasted garbage when you press Enter. • Background task completion messages no longer grow unbounded when the task produced a huge log. • Plan viewer scrollbar now responds to clicks on the border column and renders without dark stripes in • Expired external auth provider credentials now correctly trigger the interactive sign-in flow instead of a silent 401 loop. Performance: • /btw side questions now reuse the parent session’s cached prefix for faster responses. • Doctor and tmux-backed startup are now faster when no live tmux processes remain.

给关注了半年的agent team 交份作业。 这是一篇2w字长文，也是一份 Agent Team 的最佳实践。 01 一个 Agent，是怎么变成多个 Agent 的为什么真实工作会把 Task Agent 推向 Long-running Agent；为什么一个越来越好用的 Agent 最后又必须分化；多个 Agent 出现以后，瓶颈为什么会转移到 Human。 02 先让“谁负责什么”离开 Human 的脑子Agent 怎样获得稳定身份、Domain 和 Scope；Human 与 Agent 怎样查询当前 Team 的责任结构；真实协作又怎样逐渐沉淀为 Organization 与 Collaboration。 03 让 Agent 自己开始协作Agent Message 带来了什么变化；一个 Agent 怎样把工作直接交给另一个 Agent；为什么 Agent 沟通不是一次把一切说完。 04 Message 负责沟通，Topic 负责收口当工作跨越多个 Agent、多个 Turn 和多天以后，怎样保留唯一的当前版本；Responsible、Participant、Artifact 和 Needs You 分别解决什么问题。 05 Overview：让一支持续变化的 Agent Team 变得可治理当 Human 不再阅读每个 Agent 的全部过程，怎样从更高层观察 Team；怎样发现等待、积压和瓶颈候选；为什么治理关注的是工作流动，而不是让每个 Agent 看起来都很忙。 06 Agent Team 怎样进入真实的外部关系为什么把 Agent 接入 Slack、飞书还远远不够；Agent 对外以后，身份、角色、信任、权限和现实后果为什么必须被分别治理。 07 CodexLoom 在织什么Multiple Agents 与 Agent Team 的真正分水岭是什么；Human 在 Team 中的新位置是什么；为什么需要稳定的 Agent 和动态的 Team。

Our first Kite demo was ChatGPT ordering Uber Eats for you. It never got past the login screen. Uber's CAPTCHA was built for humans, and agents cannot pass it. On @gracegongGG's Venture with Grace, our Co-Founder & CEO @ChiZhangData traced how that moment shaped the way we design Kite. ▷ The demo died at the login, not at ordering or paying. The blocker was proving the agent was allowed to be there at all. Spinning up a browser extension to click through the CAPTCHA works, but it is a temporary stage of how agents and the internet should work together. ▷ The permanent path is identity. Every agent needs to prove who it is and where it comes from, carrying who its human or organization principal is, what its purpose is, and the exact authorization scope for that session or request. An agent showing up with no reason is either hallucinating or attacking you. ▷ Where this is heading: an API-first agentic internet, where headless agents authenticate without human login credentials. x402 and MCP are already marching toward that. Kite Agent Passport is that identity layer: verifiable identity, delegated authority, and permissions scoped per session. No more "prove you're human." 🪁