[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-ai-agent-tooling-docs-into-workflows-en":3,"article-related-ai-agent-tooling-docs-into-workflows-en":30,"series-tools-93d495f8-e764-4dd2-9c49-946f1f91c909":79},{"id":4,"slug":5,"title":6,"content":7,"summary":8,"source":9,"source_url":10,"author":11,"image_url":12,"cover_image":12,"category":13,"language":14,"translated_content":11,"related_article_id":15,"keywords":16,"key_takeaways":22,"views":26,"created_at":27,"published_at":28,"topic_cluster_id":29},"93d495f8-e764-4dd2-9c49-946f1f91c909","ai-agent-tooling-docs-into-workflows-en","AI Agent tooling that turns docs into workflows","\u003Cp data-speakable=\"summary\">This turns a static agent stack into a copy-ready workflow you can actually ship.\u003C\u002Fp>\u003Cp>I've been tracking \u003Ca href=\"\u002Ftag\u002Fai-agent\">AI agent\u003C\u002Fa> tools for a while now, and honestly, most of them feel like demos wearing production clothes. They promise a lot, then you open the thing and it’s either a wrapper around a chat box or a pile of settings that look impressive until you try to do real work. I keep running into the same frustration: the tool can talk about agents, but it can’t help me coordinate them. It can’t make the terminal, the editor, the protocol layer, and the team workflow feel like one system. So when I saw a write-up on \u003Ca href=\"https:\u002F\u002Fzhuanlan.zhihu.com\u002Fp\u002F2061056409358296138\">Zhuanlan Zhihu\u003C\u002Fa> about an AI agent tooling stack built around Rust, an agent panel, ACP, and real-time collaboration, I paid attention. That combination is the part people usually hand-wave away, and it’s exactly where the pain lives.\u003C\u002Fp>\u003Cp>The source post is on \u003Ca href=\"https:\u002F\u002Fzhuanlan.zhihu.com\u002Fp\u002F2061056409358296138\">zhuanlan.zhihu.com\u003C\u002Fa>, and it describes the stack in four blunt bullets: Rust for performance, an Agent Panel for terminal agents like Claude Agent and Codex CLI, ACP for running \u003Ca href=\"\u002Fnews\u002Fclaude-code-terminal-workflow-template-en\">Claude Code\u003C\u002Fa> internally, and real-time multi-user editing that feels closer to \u003Ca href=\"\u002Ftag\u002Fgoogle\">Google\u003C\u002Fa> Docs than a solo coding tool. I’m not quoting any view or star numbers here because the source doesn’t provide them. What I can do is break down what those claims actually mean when you’re trying to build something useful instead of just impressive.\u003C\u002Fp>\u003Ch2>Rust is not the headline; it is the apology for bad startup times\u003C\u002Fh2>\u003Cblockquote>技术架构： - Rust 构建：极致性能，启动速度极快\u003C\u002Fblockquote>\u003Cp>What this actually means is that the author is trying to solve the first thing that kills agent tools in practice: waiting. If a tool takes forever to boot, people stop using it for small tasks. Then it becomes the thing you launch only when you already know you need it, which is the opposite of how agent workflows should work.\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1784707449888-f5dh.png\" alt=\"AI Agent tooling that turns docs into workflows\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>I’ve run into this with Electron-heavy dev tools over and over. They look polished, then you wait three seconds, then five, then you start avoiding them. Agent tools make this worse because they’re already asking for trust. If the app feels sluggish before it even does anything intelligent, I’m out.\u003C\u002Fp>\u003Cp>Rust here is less about ideology and more about keeping the interaction loop tight. Fast startup, low overhead, fewer excuses. That matters because agent workflows are interrupt-driven. You open the panel, ask for help, inspect the result, tweak the prompt, run again. If each loop is slow, the whole thing collapses into friction.\u003C\u002Fp>\u003Cp>How to apply it: if you’re building or evaluating an agent tool, test the cold start first. Not the \u003Ca href=\"\u002Ftag\u002Fbenchmark\">benchmark\u003C\u002Fa>. Not the marketing line. The cold start. Measure how long it takes from click to usable UI, and then measure how long it takes to run the first agent action. If the tool can’t feel instant enough for repeated use, no amount of model branding will save it.\u003C\u002Fp>\u003Cul>\u003Cli>Prefer native performance where the workflow is interactive.\u003C\u002Fli>\u003Cli>Keep the first meaningful action under a few seconds if you can.\u003C\u002Fli>\u003Cli>Design for repeated short loops, not one giant “run the agent” moment.\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>An Agent Panel is just a nicer way to admit the terminal still matters\u003C\u002Fh2>\u003Cblockquote>Agent Panel：内置终端 Agent，支持 Claude Agent、Codex CLI 等\u003C\u002Fblockquote>\u003Cp>This part is the most practical claim in the whole post. The author is not saying “replace the terminal.” They’re saying put agent access where developers already work, and let it speak the same language as the tools we already trust. \u003Ca href=\"\u002Ftag\u002Fclaude\">Claude\u003C\u002Fa> Agent and Codex CLI are both examples of agentic command-line workflows, and the panel idea is basically a control surface for them.\u003C\u002Fp>\u003Cp>What this actually means is that the UI is not the product. The workflow is. A panel can expose multiple agents, show their state, and let you switch between them without making you context-switch into a separate app. That’s a lot more useful than a chat window that pretends to be an IDE.\u003C\u002Fp>\u003Cp>I like this direction because it respects how developers actually operate. We jump between terminal, editor, logs, and browser. A panel that sits beside those tools can become the coordination layer. It can show what the agent is doing, what command it ran, what failed, and what needs approval.\u003C\u002Fp>\u003Cp>I ran into this exact pattern when trying to use agent tools for repo cleanup. The chat interface was fine for brainstorming, but the moment I needed to inspect files, run commands, and check diffs, I wanted the terminal back. A panel solves that by keeping the agent close to execution instead of trapped in a conversation box.\u003C\u002Fp>\u003Cp>How to apply it: if you’re designing an agent panel, don’t start with “chat.” Start with actions. Show commands, outputs, approvals, errors, and file changes. Then add chat as commentary on top. If you’re a user, look for tools that let you bind agents to real execution surfaces like terminals, shells, and repo operations instead of only text prompts.\u003C\u002Fp>\u003Cul>\u003Cli>Expose tool output, not just model text.\u003C\u002Fli>\u003Cli>Keep approval boundaries visible.\u003C\u002Fli>\u003Cli>Let the user jump from suggestion to execution without re-entering context.\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>ACP sounds abstract until you realize it is about keeping Claude Code inside the machine\u003C\u002Fh2>\u003Cblockquote>ACP 协议：Agent Context Protocol，可运行 Claude Code 内部\u003C\u002Fblockquote>\u003Cp>What this actually means is that the author wants a protocol layer for agent context, so \u003Ca href=\"\u002Ftag\u002Fclaude-code\">Claude Code\u003C\u002Fa> can run inside the system instead of being bolted on from the outside. That’s a big deal because agent tools get messy when context is scattered across chat history, local files, terminal output, and external services with no shared structure.\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1784707461175-ja4n.png\" alt=\"AI Agent tooling that turns docs into workflows\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>I’ve been burned by this more times than I want to admit. One tool knows the prompt. Another tool knows the repo. A third tool knows the approval state. Then everything breaks the moment you switch tasks. A protocol is the boring answer, and boring answers are usually the right ones in developer tooling.\u003C\u002Fp>\u003Cp>ACP, as described here, is basically an attempt to standardize how agent context moves through the system. If it works, then the tool can treat agent state as something structured rather than improvised. That makes it easier to resume sessions, inspect what happened, and potentially plug in different agents without rebuilding the entire stack each time.\u003C\u002Fp>\u003Cp>How to apply it: define a context contract before you build the UI. Ask what the agent needs to know, what it’s allowed to change, what it should persist, and how another agent could read that state later. If you’re evaluating a product, ask whether context is exportable, inspectable, and reusable. If the answer is no, you’re looking at a fancy scratchpad.\u003C\u002Fp>\u003Cp>Useful reference points here are \u003Ca href=\"https:\u002F\u002Fwww.anthropic.com\u002Fclaude-code\">Claude Code\u003C\u002Fa>, \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fopenai\u002Fcodex\">Codex CLI\u003C\u002Fa>, and the broader idea of structured tool context that keeps agents from hallucinating their own working memory. The exact protocol name in the post may be specific to that project, but the design instinct is familiar: make context a first-class object, not a side effect.\u003C\u002Fp>\u003Ch2>Real-time collaboration is where solo-agent tools stop being toys\u003C\u002Fh2>\u003Cblockquote>实时多人协作：类似 Google Docs 的实时协作编辑\u003C\u002Fblockquote>\u003Cp>This is the part that actually changes the shape of the product. A lot of agent tools assume one user, one session, one brain. That works until you try to use them in a team. Then you get duplicated prompts, conflicting edits, and a lot of “wait, who changed this?” messages.\u003C\u002Fp>\u003Cp>What this actually means is that the tool is trying to make agent work observable and shared. The Google Docs comparison matters because it implies live state, concurrent editing, and visible presence. That’s not a cosmetic feature. It’s how you stop agent work from becoming private chaos.\u003C\u002Fp>\u003Cp>I’ve seen teams try to coordinate AI-assisted work through screenshots and copied prompts. It’s awful. Nobody knows which instruction produced which result, and nobody wants to be the person who replays the whole thing. Real-time collaboration gives you a shared control room. It lets one person steer while others observe, edit, or approve in the same session.\u003C\u002Fp>\u003Cp>How to apply it: if you’re building for teams, make session state shared by default. Show who is editing, who approved what, and what changed since the last turn. If you’re adopting a tool, ask whether multiple people can inspect the same agent session without exporting a pile of screenshots and logs into Slack.\u003C\u002Fp>\u003Cul>\u003Cli>Shared sessions beat private prompt archaeology.\u003C\u002Fli>\u003Cli>Presence indicators matter when decisions happen live.\u003C\u002Fli>\u003Cli>Audit trails are not optional once multiple people touch the same agent flow.\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>The real stack is not four features; it is one loop\u003C\u002Fh2>\u003Cp>When I put these pieces together, the pattern gets clearer. Rust handles speed. The Agent Panel handles access. ACP handles context. Real-time collaboration handles team use. None of those matters alone if the others are missing.\u003C\u002Fp>\u003Cp>That’s the part people miss when they read feature lists. They treat each line as a checkbox. I don’t think that’s the right way to read this post. I think it is describing a workflow loop: open fast, act locally, keep context structured, and let other people join the same session without recreating the state by hand.\u003C\u002Fp>\u003Cp>This is also where a lot of agent products fail. They overinvest in the model hookup and underinvest in the surrounding machinery. But the surrounding machinery is what makes the model usable every day. If the stack cannot preserve context and collaboration, then every task becomes a fresh negotiation with the tool.\u003C\u002Fp>\u003Cp>How to apply it: map your current workflow as a loop. Where do you wait? Where do you lose context? Where do you hand work to another person? Then fix those breaks in order. Start with speed, then context, then shared editing. Don’t try to bolt collaboration on top of a slow, stateful mess and hope it behaves.\u003C\u002Fp>\u003Ch2>If I were shipping this, I would make the defaults boring and strict\u003C\u002Fh2>\u003Cp>Here’s my opinionated version: agent tools should be boring in the best way. They should open fast, keep state visible, avoid hidden magic, and make it easy to tell what the agent did versus what a human changed. The more important the tool becomes, the less I want mystery meat behavior.\u003C\u002Fp>\u003Cp>The Zhihu post is interesting because it points toward that discipline. It doesn’t read like a “look at this amazing demo” pitch. It reads like someone trying to assemble the missing pieces of a real working environment for agents. That’s a much better direction.\u003C\u002Fp>\u003Cp>How to apply it: write down the invariants your tool should never violate. Examples: no hidden context mutation, no unlogged agent action, no shared edit without attribution, no cold start that makes people quit before first use. Then make those rules visible in the product.\u003C\u002Fp>\u003Cp>If you’re a developer choosing tools, use those same rules as your filter. A flashy agent that cannot explain itself is a liability. A plain tool that keeps state clean is worth more than a hundred polished screenshots.\u003C\u002Fp>\u003Ch2>The template you can copy\u003C\u002Fh2>\u003Cpre>\u003Ccode>## AI agent workflow spec\u003C\u002Fcode>\u003C\u002Fpre>\u003Cpre>\u003Ccode>Goal: build an agent workflow that is fast, inspectable, and team-friendly.\u003C\u002Fcode>\u003C\u002Fpre>\u003Cpre>\u003Ccode>### 1) Startup\u003C\u002Fcode>\u003C\u002Fpre>\u003Cpre>\u003Ccode>- Use a native runtime where possible.\u003C\u002Fcode>\u003C\u002Fpre>\u003Cpre>\u003Ccode>- Target a cold start that feels immediate to the user.\u003C\u002Fcode>\u003C\u002Fpre>\u003Cpre>\u003Ccode>- Measure click-to-usable time and first-action time.\u003C\u002Fcode>\u003C\u002Fpre>\u003Cpre>\u003Ccode>### 2) Agent access\u003C\u002Fcode>\u003C\u002Fpre>\u003Cpre>\u003Ccode>- Expose agents through a panel beside the terminal\u002Feditor.\u003C\u002Fcode>\u003C\u002Fpre>\u003Cpre>\u003Ccode>- Show commands, outputs, approvals, and failures.\u003C\u002Fcode>\u003C\u002Fpre>\u003Cpre>\u003Ccode>- Support multiple agent entry points, such as CLI and panel actions.\u003C\u002Fcode>\u003C\u002Fpre>\u003Cpre>\u003Ccode>### 3) Context model\u003C\u002Fcode>\u003C\u002Fpre>\u003Cpre>\u003Ccode>- Define a structured context object.\u003C\u002Fcode>\u003C\u002Fpre>\u003Cpre>\u003Ccode>- Store: task goal, repo state, approvals, tool outputs, and session history.\u003C\u002Fcode>\u003C\u002Fpre>\u003Cpre>\u003Ccode>- Make context readable, exportable, and resumable.\u003C\u002Fcode>\u003C\u002Fpre>\u003Cpre>\u003Ccode>### 4) Collaboration\u003C\u002Fcode>\u003C\u002Fpre>\u003Cpre>\u003Ccode>- Support shared sessions with live presence.\u003C\u002Fcode>\u003C\u002Fpre>\u003Cpre>\u003Ccode>- Track who edited, approved, or executed each step.\u003C\u002Fcode>\u003C\u002Fpre>\u003Cpre>\u003Ccode>- Keep an audit trail of agent actions and human overrides.\u003C\u002Fcode>\u003C\u002Fpre>\u003Cpre>\u003Ccode>### 5) Guardrails\u003C\u002Fcode>\u003C\u002Fpre>\u003Cpre>\u003Ccode>- No hidden state changes.\u003C\u002Fcode>\u003C\u002Fpre>\u003Cpre>\u003Ccode>- No unlogged tool execution.\u003C\u002Fcode>\u003C\u002Fpre>\u003Cpre>\u003Ccode>- No shared edit without attribution.\u003C\u002Fcode>\u003C\u002Fpre>\u003Cpre>\u003Ccode>### 6) Evaluation checklist\u003C\u002Fcode>\u003C\u002Fpre>\u003Cpre>\u003Ccode>- Does it open fast enough to use repeatedly?\u003C\u002Fcode>\u003C\u002Fpre>\u003Cpre>\u003Ccode>- Can I see what the agent did?\u003C\u002Fcode>\u003C\u002Fpre>\u003Cpre>\u003Ccode>- Can another person join the same session without rebuilding context?\u003C\u002Fcode>\u003C\u002Fpre>\u003Cpre>\u003Ccode>- Can I resume the work later without guessing?\u003C\u002Fcode>\u003C\u002Fpre>\u003Cpre>\u003Ccode>### 7) Copy-ready prompt for the team\u003C\u002Fcode>\u003C\u002Fpre>\u003Cpre>\u003Ccode>You are working inside a shared AI agent workflow.\n\nRules:\n1. Keep all context structured.\n2. Log every action and output.\n3. Ask before destructive changes.\n4. Prefer short, inspectable steps.\n5. Make it easy for another person to resume this session.\n\nTask:\n[Describe the task]\n\nContext:\n[Paste repo state, constraints, and approvals]\n\nOutput format:\n- Plan\n- Actions taken\n- Results\n- Open questions\n- Next step\u003C\u002Fcode>\u003C\u002Fpre>\u003Cp>Source attribution: I broke this down from the Zhihu post at \u003Ca href=\"https:\u002F\u002Fzhuanlan.zhihu.com\u002Fp\u002F2061056409358296138\">https:\u002F\u002Fzhuanlan.zhihu.com\u002Fp\u002F2061056409358296138\u003C\u002Fa>. The structure and template above are my own interpretation of the ideas in that source, not a verbatim reproduction.\u003C\u002Fp>","A practical breakdown of one AI agent stack with a copy-ready template for building faster workflows.","zhuanlan.zhihu.com","https:\u002F\u002Fzhuanlan.zhihu.com\u002Fp\u002F2061056409358296138",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1784707449888-f5dh.png","tools","en","0ef9f5bc-c841-4f67-be97-3bb5fd7ec655",[17,18,19,20,21],"AI agents","Rust","Claude Code","protocols","collaboration",[23,24,25],"Fast startup matters because agent tools live or die on repeated short loops.","A panel is more useful than a chat box when it exposes real execution and approvals.","Structured context and shared sessions are the difference between demos and team workflows.",1,"2026-07-22T08:03:27.861717+00:00","2026-07-22T08:03:27.855+00:00","a7343b93-37cc-4634-a2bc-707f6275bdb6",{"tags":31,"relatedLang":38,"relatedPosts":42},[32,34,36],{"name":18,"slug":33},"rust",{"name":19,"slug":35},"claude-code",{"name":17,"slug":37},"ai-agents",{"id":15,"slug":39,"title":40,"language":41},"ai-agent-tool-landscape-report-2026-07-zh","AI Agent 工具選型拆成四層","zh",[43,49,55,61,67,73],{"id":44,"slug":45,"title":46,"cover_image":47,"image_url":47,"created_at":48,"category":13},"b29c9a4b-0350-4171-9c68-cd8c633b7593","ai-rewrites-postgresql-in-rust-en","How AI Rewrites PostgreSQL in Rust","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1784709243935-mcwk.png","2026-07-22T08:33:28.280911+00:00",{"id":50,"slug":51,"title":52,"cover_image":53,"image_url":53,"created_at":54,"category":13},"41d705ff-ae43-457b-be0c-626acf99abfd","docker-desktop-windows-install-flow-en","Docker Desktop on Windows in one clean install flow","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1784656997287-5hio.png","2026-07-21T18:02:37.402109+00:00",{"id":56,"slug":57,"title":58,"cover_image":59,"image_url":59,"created_at":60,"category":13},"24569438-14fd-4295-99a5-145c6846b030","grok-4-5-cursor-launch-benchmarks-pricing-en","Grok 4.5 hits Cursor with $2\u002F$6 pricing","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1784536369476-at9z.png","2026-07-20T08:32:24.692259+00:00",{"id":62,"slug":63,"title":64,"cover_image":65,"image_url":65,"created_at":66,"category":13},"6bc7f633-a474-417a-9dbe-d4fbc46e3e61","agt-governed-agent-calls-en","AGT turns agent calls into governed actions","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1784405021642-7891.png","2026-07-18T20:03:17.886807+00:00",{"id":68,"slug":69,"title":70,"cover_image":71,"image_url":71,"created_at":72,"category":13},"7f260a96-7ecb-4379-8fa4-ea0854c1fa06","openclaw-v2026-7-1-control-ui-workspace-en","OpenClaw v2026.7.1 turns control UI into a workspace","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1784358239202-kzmg.png","2026-07-18T07:03:26.009425+00:00",{"id":74,"slug":75,"title":76,"cover_image":77,"image_url":77,"created_at":78,"category":13},"9d707772-c135-4706-889f-e36ca09cd4ae","openai-screenless-speaker-turns-chatgpt-companion-en","OpenAI’s screenless speaker turns ChatGPT into a companion","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1784289805829-eiuq.png","2026-07-17T12:02:53.86563+00:00",[80,85,90,95,100,105,110,115,120,125],{"id":81,"slug":82,"title":83,"created_at":84},"8008f1a9-7a00-4bad-88c9-3eedc9c6b4b1","surepath-ai-mcp-policy-controls-en","SurePath AI's New MCP Policy Controls Enhance AI Security","2026-03-26T01:26:52.222015+00:00",{"id":86,"slug":87,"title":88,"created_at":89},"27e39a8f-b65d-4f7b-a875-859e2b210156","mcp-standard-ai-tools-2026-en","MCP Standard in 2026: Integrating AI Tools","2026-03-26T01:27:43.127519+00:00",{"id":91,"slug":92,"title":93,"created_at":94},"165f9a19-c92d-46ba-b3f0-7125f662921d","rag-2026-transforming-enterprise-ai-en","How RAG in 2026 is Transforming Enterprise AI","2026-03-26T01:28:11.485236+00:00",{"id":96,"slug":97,"title":98,"created_at":99},"6a2a8e6e-b956-49d8-be12-cc47bdc132b2","mastering-ai-prompts-2026-guide-en","Mastering AI Prompts: A 2026 Guide for Developers","2026-03-26T01:29:07.835148+00:00",{"id":101,"slug":102,"title":103,"created_at":104},"3ab2c67e-4664-4c67-a013-687a2f605814","garry-tan-open-sources-claude-code-toolkit-en","Garry Tan Open-Sources a Claude Code Toolkit","2026-03-26T08:26:20.245934+00:00",{"id":106,"slug":107,"title":108,"created_at":109},"66a7cbf8-7e76-41d4-9bbf-eaca9761bf69","github-ai-projects-to-watch-in-2026-en","20 GitHub AI Projects to Watch in 2026","2026-03-26T08:28:09.752027+00:00",{"id":111,"slug":112,"title":113,"created_at":114},"9f332fda-eace-448a-a292-2283951eee71","practical-github-guide-learning-ml-2026-en","A Practical GitHub Guide to Learning ML in 2026","2026-03-27T01:16:50.125678+00:00",{"id":116,"slug":117,"title":118,"created_at":119},"1b1f637d-0f4d-42bd-974b-07b53829144d","aiml-2026-student-ai-ml-lab-repo-review-en","AIML-2026 Is a Bare-Bones Student Lab Repo","2026-03-27T01:21:51.661231+00:00",{"id":121,"slug":122,"title":123,"created_at":124},"6d1bf3f6-e191-4d30-b55b-8a0722fa6afe","ai-trending-github-repos-and-research-feeds-en","AI Trending Tracks Repos and Research Feeds","2026-03-27T01:31:35.709532+00:00",{"id":126,"slug":127,"title":128,"created_at":129},"010539a1-4c3a-4bd3-937a-26616422ee0d","awesome-ai-for-science-research-tools-map-en","Awesome AI for Science Is Becoming a Real Research Map","2026-03-27T01:46:50.89513+00:00"]