[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-xai-grok-build-open-source":3,"topics-all":33,"news-related-95b38793-01e2-48d7-a02b-24bf02b2fed5":52},{"id":4,"title":5,"summary":6,"content":6,"original_url":7,"source_id":8,"tags":9,"translations":20,"news_slug":26,"published_at":27,"created_at":28,"modified_at":29,"is_published":30,"publish_type":31,"image_url":13,"view_count":32},"95b38793-01e2-48d7-a02b-24bf02b2fed5","xAI 把 Grok Build 全栈开源：刚被曝偷传完整代码库的 coding agent，如何用 Apache 2.0 救场","今天,xAI 在 GitHub 上以 Apache 2.0 协议公开了自家编程智能体 Grok Build 的完整源码。仓库（github.com\u002Fxai-org\u002Fgrok-build）几天内已经收获 10.7k Star、1.8k Fork,整套 CLI\u002FTUI 与 agent runtime 几乎全部由 Rust（99.6%）写成,与官方二进制同步发布。\n\n这个时间点非常微妙——就在几天前,安全社区刚披露 Grok Build 在读取或处理文件时,会把整份原始内容未做任何脱敏地上传到 xAI 使用的 Google Cloud Storage,有用户报告连 SSH 私钥、密码管理器数据库都被一并打包;其数据保留时长也明显超过 Claude Code、Gemini CLI、OpenAI Codex 等同类工具。Elon Musk 公开承诺将彻底删除此前上传到服务器的所有用户数据。\n\n把闭源工具直接开源,是挽回开发者信任最干脆的做法——任何人都可以进仓库 grep 上传逻辑、审数据流,审计取代公关。xAI 同时附带 SECURITY.md 与清晰的 Rust crate 划分(xai-grok-pager 负责 TUI、xai-grok-shell 负责 agent runtime、xai-grok-tools 负责工具实现、xai-grok-workspace 负责文件系统与 VCS),开发者既能跑 cargo check 快速校验,也能读 Cargo.toml 看清楚依赖闭包。安装一行 `curl -fsSL https:\u002F\u002Fx.ai\u002Fcli\u002Finstall.sh | bash` 即可,headless 模式还能直接接进 CI\u002FCD。\n\nGrok Build 现已切换到 Grok 4.5 提供推理,原生支持 subagent 并行、Plan Mode、MCP server、AGENTS.md,与 Claude Code、Cursor 正面竞争。\n\n闭源阵营把 AI 编程工具变成黑盒,开源阵营则把代码摊给社区审查;这一轮 coding agent 的真正分水岭,不是哪家模型更强,而是哪家愿意先把自己切开。","https:\u002F\u002Fgithub.com\u002Fxai-org\u002Fgrok-build","b82e17a3-1dbd-4b5d-88dc-9f518f917cc0",[10,14,17],{"id":11,"name":12,"slug":12,"description":13,"color":13},"1fcfaaf2-67de-43d3-9e35-5784852fec60","ai-safety",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"e82b2d09-81b2-43d1-977e-e018443b3c14","coding-agent",{"id":18,"name":19,"slug":19,"description":13,"color":13},"b9bd9039-fcdb-41a8-b85b-fc1587def2b9","open-source",[21],{"id":22,"lang":23,"title":24,"summary":25,"content":13},"3cd23279-1eb3-4cd1-8d07-93065b429027","en","xAI open-sources Grok Build after upload controversy","Today, xAI released the complete source code of its programming agent Grok Build on GitHub under Apache 2.0. The repository (github.com\u002Fxai-org\u002Fgrok-build) has racked up 10.7k Stars and 1.8k Forks in just a few days; the entire CLI\u002FTUI and agent runtime is written almost entirely in Rust (99.6%), and ships in sync with the official binary. The timing is delicate — only days earlier, the security community disclosed that when Grok Build reads or processes files, it uploads the entire raw content to Google Cloud Storage used by xAI with no redaction, and users have reported that even SSH private keys and password-manager databases were bundled in; its data retention period was also clearly longer than comparable tools like Claude Code, Gemini CLI, and OpenAI Codex. Elon Musk publicly pledged to completely delete all previously uploaded user data from the servers. Open-sourcing a closed-source tool is the most straightforward way to win back developer trust — anyone can grep the upload logic in the repo, audit the data flow, and replace PR with audit. xAI also includes a SECURITY.md and clear Rust crate separation (xai-grok-pager for TUI, xai-grok-shell for agent runtime, xai-grok-tools for tool implementation, xai-grok-workspace for filesystem and VCS), so developers can run `cargo check` for quick validation, or read Cargo.toml for a clear view of the dependency closure. Installation is one line: `curl -fsSL https:\u002F\u002Fx.ai\u002Fcli\u002Finstall.sh | bash`; headless mode can plug directly into CI\u002FCD. Grok Build has now switched to Grok 4.5 for inference, with native support for subagent parallelism, Plan Mode, MCP server, AGENTS.md — competing head-on with Claude Code and Cursor. The closed-source camp turns AI programming tools into black boxes; the open-source camp lays the code out for community audit. The real watershed for this round of coding agents isn't whose model is stronger, but which company is willing to open itself up first.","xai-grok-build-open-source","2026-07-16T16:05:00Z","2026-07-16T16:08:11.121360Z","2026-08-19T02:08:40.142862Z",true,"agent",186,[34,43],{"slug":35,"tag_slug":35,"title_zh":36,"title_en":37,"intro_zh":38,"intro_en":39,"id":40,"is_active":30,"created_at":41,"modified_at":42},"ai-for-science","AI for Science 2026：从 UniPert 到 GPT-Rosalind 的硬核进化","AI for Science 2026: from UniPert to GPT-Rosalind","生命科学、化学材料、物理世界模型——AI 正在从\"语言工具\"变成\"实验伙伴\"。本专题收录 AI 在三大科学方向的关键节点：UniPert 统一基因与化学扰动空间、GPT-Rosalind 端到端生命科学推理、达摩院 AI 智能体 28 小时找到 4 种超导新材料、Anthropic Claude Science 把工作台做成标准品。","From language tool to lab partner — AI is reshaping life sciences, chemistry\u002Fmaterials, and physical world models. This topic covers the key milestones: UniPert unifying genetic-chemical perturbation spaces, GPT-Rosalind's end-to-end life-sciences reasoning, DAMO's AI agent discovering 4 superconducting materials in 28 hours, and Anthropic's Claude Science workbench going mainstream.","988a4300-5fab-41c4-b5d8-63711a2dc757","2026-09-10T01:34:15.296649Z","2026-09-10T01:34:15.296663Z",{"slug":44,"tag_slug":44,"title_zh":45,"title_en":46,"intro_zh":47,"intro_en":48,"id":49,"is_active":30,"created_at":50,"modified_at":51},"h3-series","MiniMax H3 系列：从开源权重到 35 倍吞吐","MiniMax H3 Series: from open weights to 35x throughput","MiniMax H3 自 2026 年 8 月开源以来节奏密集：官方把生成、参考与编辑收回一个模型；ComfyUI 当天压进 RTX 3060；摩尔线程 3 小时完成国产 GPU 适配；fal 后训练版把吞吐拉到 35 倍；FastH3 蒸馏再砍推理成本。本专题持续追踪 H3 的发布—开源—蒸馏—部署全链路。","Since MiniMax open-sourced H3 in August 2026 the pace has been relentless: one unified omni-modal model, same-day ComfyUI support down to an RTX 3060, a 3-hour Day-0 port to Moore Threads GPUs, fal's post-trained H3 Max at 35x throughput, and FastH3 distillation cutting inference cost further. This topic tracks the full H3 chain — release, open weights, distillation, deployment.","83ef0daa-3c31-4cb3-86ed-e5ee58654d5f","2026-09-08T07:33:19.942193Z","2026-09-08T07:33:19.942209Z",{"items":53},[54,59,64,69,74,79],{"id":55,"title":56,"news_slug":57,"published_at":58},"9f800588-ad3b-4eee-a8a6-e2db1ac8f014","GLM-5.3:只靠后训练把 743B 基座打成新 SOTA,网络安全的\"涌现\"打了 Z.ai 一个措手不及","glm-5-3-post-training-emergent-cyber","2026-08-14T08:00:00+00:00",{"id":60,"title":61,"news_slug":62,"published_at":63},"de219584-58fc-45ad-91ea-0049a5cbcf10","OpenAI 开源 Codex Security CLI:把 AI 安全检测塞进每个 PR","openai-codex-security-cli-opensource","2026-07-29T10:30:00+00:00",{"id":65,"title":66,"news_slug":67,"published_at":68},"2222d50a-9e72-461e-b202-8f297975c297","rsync维护者回应「Claude代笔」争议：当关键基础设施遇上AI编程代理","rsync-tridge-claude-ghostwriting-infrastructure","2026-06-05T19:00:00+00:00",{"id":70,"title":71,"news_slug":72,"published_at":73},"084a216d-7f7a-405f-8a3d-218cbfe56870","开源项目「试探」AI助手：jqwik隐藏指令引发安全争议","jqwik-hidden-instruction-prompt-injection","2026-05-29T07:08:00+00:00",{"id":75,"title":76,"news_slug":77,"published_at":78},"9dd4a859-1153-4ecc-b69d-4ba4c5431129","智谱被开发者抓包后紧急上线数据零留存","zhipu-maas-zero-data-retention-zcode","2026-09-21T07:00:00+00:00",{"id":80,"title":81,"news_slug":82,"published_at":83},"fa4c43df-5d62-464b-b4b0-3a6854000644","AI 攻破 OpenAI 内网全程复盘:靠 Anthropic 模型当武器,72 小时打开自家后门","hacktron-claude-openai-monorepo-rce-72h","2026-09-21T03:00:00+00:00"]