与「REVIEW」相关的搜索结果

ChatGPT can now buy things for you. On Friday OpenAI launched ChatGPT Wallet, so its AI agents can check out with your card. What’s actually in it: → Your card number isn’t shared with the agent or the store. It pays with a substitute → You review the items and the total, then approve every purchase with Face ID or a fingerprint → No spending limits yet. OpenAI says they’re coming → For now it’s a limited rollout to some US ChatGPT Pro users That approval tap is the part to take seriously. In a Cisco and Carnegie Mellon study, AI agents picked pricier options for users they thought were rich, and a clear budget nearly closed the gap for most models. So if you try it, give the agent a number and read the total before you approve. Would you let an AI agent buy something for you? Sources: OpenAI, TechCrunch #AI# #ChatGPT#

Who decides the answer to a prediction market? Usually the creator. Or an AI on a server nobody can check. So we built Noa Court - an AI resolver for Noa, live on NEAR mainnet. In the video we ask Gemini a vague question: "will the NEAR hackathon get more than 2000 projects?" - It finds the real source (Nearhackathon API) and samples it live: 1,803 right now - Turns the vague question into precise, hashed rules - We review the market and sign with a NEAR wallet - Market #27# goes live on Noa, published with readable rules How it resolves: - One Court contract, one pinned resolver running in an @out_layer TEE - Anyone can trigger it. Nobody can pick the answer - Numbers? Decided by code - Words? 3 NEAR AI judges must all agree and quote the evidence Asked "did a crypto hack make the news?", all 3 judges quoted a headline that never says "hack". It refuses to guess: - Prompt injection - flagged - Dead or changed source - void + refund Works with prices, NEAR tokens (powered by Intear & JustHoot), news feeds, web pages, any JSON API. Plug it into any market, contract or agent. Add the MCP to Claude, ChatGPT, Gemini - any AI that speaks MCP: Code: What would you never let an AI resolve? Built at NEAR HACKATHON @NEARProtocol @NearBuilders #NEARHackathon#

@meetmotes: NFT-owned AI agents on Robinhood Chain. You check your crypto balance regularly. But when did you last check your wallet approvals? That question comes from a creator thread retweeted by FINOMZY II, and it lands harder once an AI agent holds your keys. An approval is permission that outlives the moment you gave it. A session key you forgot is still a key. Motes answers with structure, not reassurance. The agent works through a session key that is constrained, rotatable and revocable by the owner, and invalidated when ownership changes. The owner key stays the high authority: change permissions, stop the service, withdraw. The docs build the check into the design. Every call is validated against the owner-signed policy, and the weekly ledger shows what the agent did. Review approvals the way you review balances. An agent you can revoke is an agent you can trust. My verdict: in agent finance, the revoke button is the product.

My landlord raised my rent by $400 a month and gave me 30 days to accept or move out. His text said: "Market rate. Nothing personal." I replied, "Understood. Can you send the new lease so I can review it?" He sent it within the hour. I read all 11 pages twice, and on page 9, I found one line he had clearly forgotten was there...

资深工程师写代码前后有一套流程：先写 spec，再拆任务，测过才算数，合并前过一道 review。现在这套流程被打包成了 Skill，可以直接装给 Claude Code 这类写代码的 agent。 仓库叫 agent-skills，作者是 Addy Osmani。这周涨了 3502 星，总数到了 11.7 万，MIT 许可。 它把开发拆成六段：定义、计划、构建、验证、评审、上线，每段都有对应的斜杠命令。/spec 先把要做什么写清楚再动代码；/plan 把活切成很小的原子任务；/build 一次只做一片；/test 拿测试当证据，跑过了才算完成。 另外几个也实用：/webperf 先测量再优化，/code-simplify 要的是写得清楚，不是写得巧。最值钱的是 /constraints，质量标准定一次，之后每一步都按它卡。 按作者的说法，这些 Skill 记的是资深工程师的工作流和质量关卡，打包起来是为了让 agent 每个阶段都稳定照做。 agent 最常见的毛病就是一上来就写代码，一口气改一大片。这套东西正好反过来，把动手前和合并前那几步补了回去。 在 Claude Code 里它们是作为插件命令装进去的。/review 和自带命令重名，用的时候要带上命名空间。

腾讯这波操作，等于直接给 Agent 组建了一支完整的研发班子！ 才刚开源的 LoopForge，能把一项开发任务切分成一整套端到端的研发流水线： 头一步先把需求定下来，再把技术方案敲定 写代码这件事交由 AI Agent 来完成 再单独派一个 Agent 去做 Code Review 测试跑一轮，要是查出毛病就退回前序环节返工 交付过程全程留痕，哪怕中途断了也能续上接着干 更狠的是，这整条流程能沉淀成一份 SOP，往后遇到相似的功能开发，拿出来套用即可，省得每次都从零折腾一遍 Claude Code、Codex、CodeBuddy 三样都支持。个人自用的话，装个 Local 版就够，用不着把整套平台部署起来。 项目完全开源，走的 MIT 协议。 GitHub：

今天 GitHub 被「浏览器三巨头 + 安全扫描 + 代码审查」大厂联军屠榜了！ 5 个星标暴增最狠的项目，专业拆解下！ 1. GoogleChrome/modern-web-guidance-src Modern Web Guidance 的源码仓：把 Web 平台最佳实践和浏览器兼容数据做成技能，直接塞进你的编程 Agent，让它别再写过时套路；一行 npx 就能安装、搜索指南、扫描旧代码。由 Chrome、Edge、Firefox 和 MDN 团队共同维护。 🔗 直达： AI 写前端最大的毛病就是「十年前的 Stack Overflow 味」，这回浏览器厂商亲自下场补课。1,875 星一天，含金量拉满！ 🟢 2. openai/codex-security OpenAI 官方的 Codex Security 命令行和 TypeScript SDK：扫描仓库或 Git 改动、验证漏洞、生成补丁并复核修复，还能导出 SARIF/JSON/CSV，接 Linear、Jira、GitHub 和 CI。 🔗 直达： 找漏洞、验漏洞、修漏洞一条龙，安全工程师多了个不下班的搭档。754 星一天，总星 1.3 万+，AppSec 正式进入 Agent 时代！ 🟢🟢 3. tailscale/tailcat Tailscale 官方的「不用 Tailscale 的 Tailscale」：像 netcat 一样在两台机器间开 WireGuard 端到端加密隧道，自动打洞直连，不需要账号、不需要 root，纯用户态命令行和 Go 库。 🔗 直达： 一个短地址就能把两台机器连起来，穿透 NAT 像喝水一样简单。324 星一天，网络老炮狂喜！ 🟢🟢🟢 4. alibaba/open-code-review 阿里内部用了两年的 AI 代码审查工具开源：读 Git diff，交给带工具调用的 Agent 去翻全文件、搜代码库，输出精确到行的结构化评审意见；README 称同模型下精确率和 F1 明显高于通用 Agent，token 只用约九分之一。 🔗 直达： 数万开发者、数百万缺陷练出来的审查官，现在人人能请。206 星一天，总星 5 万+，大厂内功直接外放！ 🟢🟢🟢🟢 5. ChromeDevTools/chrome-devtools-mcp Chrome DevTools 官方给 Agent 的接口：MCP 服务 + 命令行，让 Claude、Cursor、Copilot 等编程 Agent 直接操控、检查一个真实 Chrome，录性能 trace、看网络请求、截图、读控制台。 🔗 直达： Agent 终于不再「盲写前端」，能自己打开 F12 找 bug 了。145 星一天，总星 5.5 万+，浏览器调试彻底交给 AI！ 🟢🟢🟢🟢🟢 ⚠️⚠️ 总结： 今天的榜单写满了一个信号：大厂开始把自家内功直接喂给 Agent（Chrome 的 Web 指南和 DevTools、OpenAI 的安全扫描、阿里的代码审查）。工具越来越像「给 AI 用的」，谁先把 Agent 武装到牙齿，谁就站在下一波生产力的 C 位。

We are deeply sorry for the disruption this incident has caused our customers and partners. We know the trust they place in our services and take that responsibility extremely seriously. We are seeing broad recovery and stability across our services. This was a technical issue not related to cyber or AI / agents. The incident was introduced by an issue during an earlier network capacity upgrade that surfaced following a recent network change. We have addressed the underlying issue and continue to restore service to remaining customers and regions. Given the scale of demand, we restored services gradually, validating stability at each stage. While this took longer, it helped ensure a stable restoration. We will provide a full post-incident report once our review is complete.