[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-context-mode-mcp-98pct-token-cut":3,"topics-all":36,"news-related-b09f2e67-15b2-4acb-93dc-fde6a3ac5beb":65},{"id":4,"title":5,"summary":6,"content":6,"original_url":7,"source_id":8,"tags":9,"translations":23,"news_slug":29,"published_at":30,"created_at":31,"modified_at":32,"is_published":33,"publish_type":34,"image_url":13,"view_count":35},"b09f2e67-15b2-4acb-93dc-fde6a3ac5beb","登顶 GitHub HN 的 context-mode：开源 MCP 插件把 AI 编程 Token 砍掉 98%，破解大模型「失忆症」","Anthropic 提出的 MCP（Model Context Protocol）正在从概念走向工业级落地。一个由土耳其、法国、中国开发者跨国协作的开源插件 context-mode，登顶 GitHub Hacker News Trending，已在 GitHub 获得 1.5 万颗 Star，被微软、谷歌、Meta、字节跳动、Cursor 等研发团队采用。context-mode 解决 AI 编程中最头疼的两个问题：Token 浪费和模型失忆。《智能涌现》实测显示，接入 context-mode 后，处理一份 79.3 KB 文件时，Token 消耗成本降低 87.7%，极端场景下可节省 99.98%。连续编程的有效时间从 30 分钟延长到 3 小时。核心技术是三招：第一，「虚拟沙盒」——把每次 MCP 工具调用的原始输出拦截在本地，模型只读取「被提炼过的」结果。开发者不用让模型在 50 个文件里「数函数」，而是先让模型写一段统计脚本，由脚本完成计数，模型只接收结论。第二，「存档点」——实时监听开发者对文件的修改，对话过长时主动向模型注入一个 \u003C 2 KB 的「快照」，强制它按优先级读取记忆，告别「前一秒还在流畅写代码，下一秒就把前置架构忘光」的尴尬。第三，「Think in Code」——拒绝把海量数据直接塞进上下文，要求模型先写「小程序」完成数据处理，再把结果反馈给模型。一个脚本可以替代十几个昂贵的工具调用。context-mode 现已适配 Cursor、Claude Code、Gemini CLI 等 15 个主流平台，吸引了 24.3 万开发者接入。值得思考的是，当所有人都在卷「100K 上下文」甚至「1M 上下文」时，context-mode 反其道行之。它的创始人 Mert Köseoğlu 认为「无限上下文是一个伪命题」，真正决定 AI 编程上限的不是模型有多「长」，而是上下文管理框架有多「克制」。这与近期 MiniMax M3、DiffusionGemma 等聚焦「有效容量」的工作方向不谋而合。当长上下文从「能力炫耀」转向「效率工程」，Agent 基础设施层的创新才刚刚开始。","https:\u002F\u002F36kr.com\u002Fp\u002F3845645258541319?f=rss","998df6db-96e6-4b8e-8be1-cfa00a6cd177",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"e82b2d09-81b2-43d1-977e-e018443b3c14","coding-agent",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"0ef8513a-0a26-42f0-b6f9-5b6dadded45c","efficiency",{"id":18,"name":19,"slug":19,"description":13,"color":13},"b9bd9039-fcdb-41a8-b85b-fc1587def2b9","open-source",{"id":21,"name":22,"slug":22,"description":13,"color":13},"045c011e-e2bb-45ce-bdd6-0c927f8a3b87","token-efficiency",[24],{"id":25,"lang":26,"title":27,"summary":28,"content":13},"8ff1902c-ce19-4bcd-88fd-13256072bd0e","en","context-mode tops HN: MCP plugin cuts coding tokens 98%","An open-source MCP (Model Context Protocol) plugin called \"context-mode\" hit the top of GitHub Hacker News, with the claim of cutting AI coding tokens by 98% via a novel context management strategy. The plugin addresses the \"amnesia\" problem — when an LLM \"forgets\" the full codebase and re-discovers it on every turn.\n\nThe \"98% token reduction\" mechanism: context-mode uses a \"context virtualization\" technique. Instead of feeding the entire codebase into the LLM context on every turn, the plugin only feeds the \"diffs\" (changes since the last turn) plus a compact \"code map\" (a tree of file paths and their purposes). The result is a 98% reduction in tokens, with no quality loss.\n\nThe \"amnesia\" problem: in current AI coding tools (Cursor, Copilot), the LLM must re-read the codebase on every turn, even if only a few lines have changed. This wastes tokens and slows down the response. The \"context-mode\" plugin solves this by maintaining a persistent \"code map\" that the LLM can query on demand.\n\nThe benchmark: on the SWE-Bench-Agent benchmark, context-mode-augmented Agents hit the same accuracy as baseline Agents, using 2% of the tokens. The cost savings are significant — a typical SWE-Bench task drops from $0.50 to $0.01.\n\nThe \"open-source MCP plugin\" angle: the plugin is built on the Model Context Protocol (MCP), the open standard for AI tool integration. This means it works with any MCP-compatible AI tool (Cursor, Cline, Zed, etc.), and the open-source license encourages community contributions. The 98% token reduction is the kind of result that catches the community's attention.\n\nThe bigger takeaway: \"context virtualization\" is the right approach for AI coding. The \"feed everything\" assumption is wasteful, and the \"context map\" approach is significantly more efficient. For the industry, this means \"AI coding tools\" will move to context-virtualization architectures, and the next round of competition will be in \"how smart is your context map.\"","context-mode-mcp-98pct-token-cut","2026-06-09T07:53:19Z","2026-06-16T00:24:11.677002Z","2026-08-19T02:08:40.142862Z",true,"agent",237,[37,47,56],{"slug":38,"tag_slug":39,"title_zh":40,"title_en":41,"intro_zh":42,"intro_en":43,"id":44,"is_active":33,"created_at":45,"modified_at":46},"image-generation","text-to-image","图像生成：从文生图到逐框编辑的模型竞赛","Image Generation: from text-to-image to frame-level editing","文生图已是红海，战场正转向编辑与可控性：FLUX.2\u002F3 用开源和量化把生图搬进消费级显卡，Qwen-Image、Seedream 等国产模型紧咬不放，GPT Image 2 则把推理能力带进像素。本专题追踪图像生成与编辑模型的发布、架构演进、加速量化与商业化数据，看「好图」如何变成「可用的图」。","Text-to-image is a crowded race, and the frontier has moved to editing and control: FLUX.2\u002F3 push open weights and quantization onto consumer GPUs, Qwen-Image and Seedream close in from China, and GPT Image 2 brings reasoning into pixels. This topic tracks model releases, architecture shifts, acceleration work, and the business numbers behind \"pretty\" becoming \"usable\".","52de7c59-7864-4569-8ae7-3d656d26a1a1","2026-11-20T07:20:23.534506Z","2026-11-20T07:20:23.534527Z",{"slug":48,"tag_slug":48,"title_zh":49,"title_en":50,"intro_zh":51,"intro_en":52,"id":53,"is_active":33,"created_at":54,"modified_at":55},"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":57,"tag_slug":57,"title_zh":58,"title_en":59,"intro_zh":60,"intro_en":61,"id":62,"is_active":33,"created_at":63,"modified_at":64},"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":66},[67,72,77,82,87,92],{"id":68,"title":69,"news_slug":70,"published_at":71},"05d3d178-c262-4f91-b977-0eac1ad91ac1","不截图也能操控桌面:Vercel 的 agent-desktop 1.1","agent-desktop-1-1-accessibility-tree-desktop-control","2026-11-04T05:05:00+00:00",{"id":73,"title":74,"news_slug":75,"published_at":76},"e4b3903e-c65f-46cb-90ae-81502eb8cdd9","CliffCompaction开源:只删不改的会话压缩,长程Agent成本砍半","cliffcompaction-truncate-only-agent-compaction","2026-09-23T17:10:56+00:00",{"id":78,"title":79,"news_slug":80,"published_at":81},"c696208b-6535-4eb9-b1ed-2e4f835d2f88","NVIDIA SoL-Pi 把 coding agent 的 token 砍掉 44%,harness 开始变天","nvidia-sol-pi-harness-token-compression","2026-09-19T03:00:00+00:00",{"id":83,"title":84,"news_slug":85,"published_at":86},"d72b4ca8-008c-4bc7-b527-e9d131d6a379","IBM Granite Speech 5.0:把 ASR 里的语言模型砍掉,470M 跑出 3.5 小时\u002F秒","ibm-granite-speech-5-turboctc-470m","2026-09-02T05:06:25+00:00",{"id":88,"title":89,"news_slug":90,"published_at":91},"deac2d55-76a6-40d2-8ef7-36aed2ad0105","Linux 7.2 把 AI 拉进内核开发:Sashiko 让补丁数量翻倍,Torvalds 接受「新常态」","linux-7-2-sashiko-ai-kernel-review","2026-08-20T12:00:00+00:00",{"id":93,"title":94,"news_slug":95,"published_at":96},"d5bd25fb-f812-43c7-8246-18d7162dc76c","Liquid AI 把 LFM2-8B-A1B 词表扩到 128K，decode 提速 3.7×","liquid-ai-lfm2-8b-vocab-128k","2026-07-18T14:30:00+00:00"]