🧠 华为放出了 openPangu 2.0 Pro:505B MoE,512K 上下文 这是一个在昇腾 NPU 上训练的超大 MoE 模型:总参数约 505B,每个 Token 激活约 18B,训练数据约 34T Tokens。模型同时提供 Thinking / Non-Thinking 模式。 架构上有几处挺有意思: • DSA + SWA 分层混合,减少长上下文推理开销 • 4 支流 mHC 残差架构 • 3 头 MTP,一次额外预测 3 个 Token • 使用 Muon 优化器训练 官方自报成绩很猛:AIME 2026 为 95.4,GPQA-Diamond 87.9,LiveCodeBench V6 85.7,SWE-bench Verified 68.5;Agent 方向也测了 MCP-Atlas、SkillsBench、Claw-Eval 等,不是只卷聊天榜。 但先别急着喊「国产开源之光」,这里有两个硬伤…

Channel
Agili
@agili_channel
On this record: Growth · Engagement · What this channel posts · Reactions · Posts · Citations · Cite this entry
32subscribers
+0 since we began measuring on 6 August 2026
Risers and fallers across the register · movement among entries of Under 1,000.
Register entry
| Telegram ID | -1003825760691 |
|---|---|
| Type | Channel |
| Username | @agili_channel |
| Description | 无形 · 无界 · 无处不在 | Formless · Boundless · Everywhere |
| Created | Between 1 February 2026 and 11 February 2026— estimated from Telegram’s id allocation, not measured. How this range is calculated. |
| First recorded | 11 August 2026 |
| Last confirmed live | 14 August 2026 |
| Measurements held | 2 |
| On Telegram | t.me/agili_channel |
Growth
| Measured (UTC) | Subscribers | Change |
|---|---|---|
| 11 Aug 2026, 02:01 | 32 | no change |
| 6 Aug 2026, 09:05 | 32 | first reading |
Engagement
20 posts held, back to 11 February 2026 — the reader has not yet reached the start of this channel’s public history, so older posts may sit further back, unread. Read across 1 pageof Telegram’s post history, 20 posts per page.
- ERR · 30 days
- 973.1%
- avg views ÷ 32 subscribers
- Avg views / post
- 311
- 5 posts measured
- Reaction rate
- —
- this channel exposes no reaction counts
- Posts in window
- 5
- of 20 held
ERR is average views per post over the last 30 days divided by subscribers, the definition TGStat uses, so this figure is comparable with the one you will see elsewhere. It falls structurally as a channel grows: a high ERR on a small channel and a low one on a large channel describe reach mathematics, not quality. We publish the figure and the sample it came from and pass no verdict on it.
ER is defined industry-wide as (forwards + reactions + comments) ÷ views— note the denominator is views, not subscribers. Telegram’s public web preview carries views and reactions but not forward or comment counts, so the reaction rate above is the reactions term only and is therefore a floor: the true ER for this channel is higher by an amount we have not measured and will not estimate.
| Window | Rolling 30 days · latest post in window 31 July 2026 |
|---|---|
| Posts held | 20 (11 February 2026 – 31 July 2026) |
| Views total | 1,557 |
| Reactions total | — |
| Forwards / comments | not exposed by the public surface — not measured, not estimated |
| Readings taken | 11 Aug 2026, 02:01 UTC |
Views are a single reading per post, taken at the time above. A post published in the last day or two is still accumulating views, which pulls the 30-day average down slightly. That is a property of the standard definition rather than a fault in it, so we keep the definition rather than “correcting” the number into something nobody can reproduce.
Precision. Telegram publishes view counts on its public widget in short form — 8.12K, 3.7M — so any reading at or above 1,000 reaches us rounded to three significant figures, and only counts below 1,000 are exact. Averages and rates derived from them are shown to the same precision rather than to the unit: a figure like 3,701,250 would assert digits nobody measured.
Reaction counts are published per emoji and rounded the same way, so a total below 1,000 is exact and a larger one is a sum that may carry a rounded component from each emoji above 1,000. Because it is a sum, it does not look rounded — read a large reaction total as three significant figures per contributing emoji rather than as the figure it prints.
What this channel posts
- Photos
- 6
- Links
- 25
Lifetime counters from Telegram’s own channel header, read 11 August 2026 — not the date at the top of this page, which is when the subscriber count was last read. Below Telegram’s rounding threshold, so these counts are exact.
Reaction mix
2 reactions across 2 posts, in 2 distinct kinds. The most used accounts for 50.0% of them.
| Reaction | Count | Share | Share, drawn |
|---|---|---|---|
| 👍 | 1 | 50.0% | |
| 🥰 | 1 | 50.0% |
No sentiment is inferred, and none should be read in. This table is ordered by count and by nothing else. Emoji do not carry stable meaning across languages or communities — 🙏 is thanks in one channel and mourning in another — so we publish which ones were pressed and how often, and pass no judgement on what an audience meant by them.
Precision. Telegram publishes reaction counts per emoji and short-forms each one — 4.34K, 1.2M — so any single kind at or above 1,000 reaches us at three significant figures, and only counts below 1,000 are exact. The shares above are ratios of those figures and carry the same error. This is also why the total here can differ slightly from a reaction total printed elsewhere on the page: both are sums of the same rounded parts, taken over samples with different edges.
Coverage. Reactions were read on 2 of the 20 sampled posts in this sample. Summed by Telegram’s own count on each post — not by adding up the per-emoji breakdown above — those same posts carry 2reactions in total: the kind of figure the paragraph above means by “a reaction total printed elsewhere on the page”.
Measured over the 20 most recent posts we hold, published 11 February 2026 to 31 July 2026, using the newest reading held for each. Telegram Stars are excluded: they are a payment, not a reaction, and they have their own section.
Recent posts
⚡️ Vercel Labs 把 TypeScript 编译成原生二进制了 scriptc 是一个 TypeScript-to-Native 编译器:普通 .ts 文件直接编译成自包含二进制,里面没有 Node、V8,也没有 JavaScript 引擎。 npm install -g scriptc scriptc build app.ts 官方给的数据挺凶: • 启动约 2.4ms,Node 约 47ms • 纯静态二进制约 170–200KB • 常见内存占用 1–4MB • 支持 async/await、class、closure、Map/Set、Buffer,以及 fs、http、https、crypto、fetch 等一批 Node API 它最聪明的设计,是不假装所有 JavaScript 都能静态编译: 1. 能确定类型和行为的代码,直接编译为 native 2. npm 依赖、any 等动态代码,显式…
⚡️ 把 GitHub Actions Runner 塞进 Cloudflare Containers,能省钱吗? 有人真跑了一遍:Worker 接收 workflow_job webhook,每个任务启动一个 ephemeral Container,跑完即销毁。冷启动到接单约 11–13 秒,这条链路确实能用。 实测有个挺反直觉的结果: • basic:1/4 vCPU、1 GiB,任务 23 秒,约 $0.00018 • lite:1/16 vCPU、256 MiB,任务 74 秒,约 $0.00015 lite 单价低,但慢了 3.2 倍,最终几乎没省钱。真要跑,basic 才是实用下限。 什么时候划算? 私有仓库已经吃完 GitHub 免费额度、手里本来就有 Workers Paid,并且存在大量 lint、通知、轻量测试这类短任务。Cloudflare 按 10ms 计费,GitHub 按分钟向上取整;10…
🧩 把 Claude Max 接进 OpenCode:Meridian 是怎么做到的? 最近发现一个挺聪明的开源项目:Meridian。 它在本机启动一个兼容 Anthropic / OpenAI API 的代理,让 OpenCode、Cline、Aider、Pi 等客户端,可以通过 Claude Max / Team 订阅调用 Claude,不需要再按 API Token 计费。 核心链路很简单: OpenCode → Meridian → Claude Agent SDK → Claude Meridian 把客户端发来的标准 API 请求翻译成官方 Claude Agent SDK 的 query() 调用,再把 SDK 的流式事件翻译回 SSE 响应。它不提取 OAuth Token、不魔改 Claude Code,也不模拟私有 API,底层仍由 Anthropic 的 SDK 负责登录、限流、缓存和上下文管理…
🎨 shadcn/ui 正式支持 React Aria 现在创建 shadcn/ui 项目时,可以在 Base UI、Radix、React Aria 三套底层实现里选了。 直接用: npx shadcn@latest init --base aria 或者在 shadcn/create 里选择 React Aria。CLI 会处理依赖、Registry、样式和组件安装,后续 shadcn add 也会自动拉取 Aria 版本。 这次不是试验性质的适配: • 所有组件都有独立文档、示例和 API 说明 • Vega、Nova、Maia 等 8 套 Style 全部支持 • React Aria 的状态选择器和依赖单独隔离 • 原有 Base UI / Radix 组件不受影响 为什么值得关注? React Aria 由 Adobe 团队长期维护,对键盘导航、焦点管理、屏幕阅读器、国际化这些「写着麻烦、漏了更麻烦」…
🦊 把 Firefox 塞进 Firefox 里,居然真跑起来了 Puter 把完整的 Firefox 编译成了 WebAssembly:Gecko 渲染引擎、浏览器 UI、SpiderMonkey JS 引擎,全都运行在当前网页的 `` 里。 直接体验: https://developer.puter.com/labs/firefox-wasm/ 几个有意思的细节: • 首次要下载约 54 MB • 网页流量通过 WISP 协议封装成 WebSocket,走 Puter 托管的代理 • 提供实验性的 JS → WASM JIT,但官方自己也写了:Expect bugs • 真能套娃:Firefox 里跑 Firefox,再在里面跑 Firefox;作者表示第三层已经很不稳定 • 作者称这次移植消耗了价值 2.5 万美元以上的 Opus / Fable Token,主要用于调试和 JIT 研究 别把它当成新的日常浏览器…
⌨️ 写 CLI,先把这份指南喂给 AI Command Line Interface Guidelines 是一份开源的现代 CLI 设计指南。它不教你选 Commander、Cobra 还是 Click,而是讨论更容易被忽略的东西:一个命令行工具到底怎样才算“好用”。 作者里有 Docker Compose 的共同创造者,内容从传统 UNIX 原则出发,但没有停留在几十年前: • 给人用时,报错要能看懂,还要告诉用户怎么修 • 给脚本用时,stdout / stderr、退出码和信号必须守规矩 • 检测 TTY 后再决定是否显示颜色、进度条和交互提示 • 支持 --help、--version、自动补全,并保持参数命名一致 • 危险操作要确认,但也得提供明确的非交互模式 • 输出别装死,也别把调试日志倒用户一脸 这份指南 2020 年首次发布时,在 Hacker News 拿到 564 分、引出 212 条讨论;Gi…
Telegram 开始做 Serverless Telegram Serverless 让 Bot 和 Mini App 的后端直接跑在 Telegram 上,不用再养 VPS、容器和 Webhook。 可以把它理解成 Telegram 原生版 Workers + D1: • JavaScript 运行在 V8 Sandbox • 内置 Bot API、SQLite 和出站 HTTP • 按 update 写 Handler,npx tgcloud push 部署 • 支持本地开发、数据库迁移和 CI 简单 Bot 的后端,可能只剩 handlers/message.js 和 schema.js。Bot Token、Webhook、数据库连接这些脏活,Telegram 全包了。 现在的问题也很明显: • 仍是小范围封闭测试,BotFather 没入口就是没资格 • 没有 npm 和文件系统,Secrets、C…
给 Agent 的临时邮箱,应该长这样 我测试了 TempMail.Best 的 Agent Skill。安装、创建邮箱、收信、读取正文和删除邮箱,整条流程都能由 Agent 完成,不需要打开网页,也不用人工接力。 邮箱可设置为 10 分钟或 1 小时有效期。邮件列表和正文通过 HTTP API 获取,Token 放在 Authorization: Bearer 中,不会出现在 URL。text 和 HTML 正文都能读取,适合验证码、确认链接、投递检查和端到端测试。 我跑了一次完整流程:创建 10 分钟邮箱,用 SendTestMail 发信,Agent 成功读到主题、发件人和正文,最后删除邮箱。创建前还需要完成 Cap proof-of-work。它不依赖浏览器环境,但保留了防滥用成本,这个设计很合理。 边界也很清楚: • 每个邮箱只保留最新 20 封邮件。 • 新邮件不会延长邮箱寿命。 • HTML 邮件属于非可…
👍1
GPT-5.6 Sol 放进真实工作流后表现怎样 Claire Vo 在 How I AI 中比较了 GPT-5.6 Sol、Terra、Luna,以及 Claude Fable 5 和 Sonnet 5。测试不是单纯跑 Benchmark,而是覆盖 PRD、原型、前端设计、调试、Agent 语气、浏览器操作和视频剪辑。 她的体验是:Fable 更擅长复杂推理,但沟通成本偏高;Sol 更像实用型执行者,做 0→1 原型、Dashboard 和开发者工具页面时,输出更完整,也更愿意做取舍。PRD 写作她反而偏 Terra,Agent 语气则认为 Sonnet 5 更自然。 这套评价标准比排行榜更接近真实开发:模型能否理解目标,把页面、代码和流程推进到可试用状态,而不是生成一份看起来完整的分析。 也别忽略限制。GPT-5.6 当时仍是有限预览,安全策略可能误拦部分正常工作;企业也开始更严格地控制 Token 成本。模型更便宜…
Yansu 会观察工作方式,再主动生成小工具 Yansu 的定位不是聊天助手。它通过桌面截图、消息应用和日常工作流识别重复模式,再生成内部工具、Dashboard、Tracker 或自动化流程。 它支持 macOS、Windows 和 Linux,可连接 Slack、Teams、飞书、Discord、Telegram、WhatsApp 和微信。记忆、对话及生成的应用以本地存储为主。后台 Computer Use 仍在开发中,目标是在虚拟层完成任务,不占用鼠标和焦点。 这个方向比“等用户写 Prompt”更进一步:很多人知道工作重复,却说不清该造什么工具。Yansu 如果能从行为中抽象需求,确实可能省下一批内部软件。 风险也同样直接。持续观察桌面和消息意味着很高的权限;生成工具的质量取决于它是否真正理解业务,而不是只做一个漂亮页面。目前公开讨论较少,仍是早期产品。 建议先接低风险流程,例如任务追踪、会议行动项或 AI 费…
Paseo 把多个 Coding Agent 接到同一个入口 Paseo 是一个开源的 Claude Code、Codex、Copilot 和 OpenCode 统一入口。它通过自建 Daemon 连接本地电脑或 VPS,再提供桌面端、移动端、Web 和 CLI。 你可以同时运行多个 Agent,用 Worktree 隔离任务,也可以让一个模型制定计划、另一个模型执行,再回收结果。手机更适合查看进度、Diff 和补充指令,真正的构建与测试仍在后端机器上完成。 这类工具解决的是上下文和设备切换,不会降低模型费用,也不会让不稳定的网络或开发环境自动变好。它最大的吸引力,是代码和执行环境仍由自己控制。 如果你经常在多台设备间切换,或已经在做多 Agent 编排,Paseo 值得试。只在单机终端里跑一个 Agent,就没必要为了统一而统一。 项目:https://paseo.sh/
Showing the 12 most recent of 20 posts we hold for @agili_channel. View and reaction counts are the latest single reading for each post, not a live figure, and a recent post is still accumulating both. A view count marked ≈ was rounded by Telegram before we ever saw it — t.me prints views in full below 1,000 and to three significant figures above, so ≈1,200,000 means somewhere between 1,150,000 and 1,249,999. Unmarked counts are exact. Text is reproduced from the public post preview and truncated for length.
Citation-graph rank
Citation-graph rank — 549,196 of 1,481,243entries in the measured graph. A weighted position computed from the forward and mention edges below — republished posts weigh more than named mentions — and recomputed periodically, over the whole graph. Published only as this ordinal position, never as a score: a position is a fact, and a score printed beside one channel’s name would read as a verdict this register does not make. The two counts beneath stay separate for the same reason mentions are never summed with forwards anywhere else on this page — a named-by count costs nothing to manufacture. The top 100 by this measure, or how it is computed.
Forward network
Republished by
Channels on the register that have forwarded this channel's posts into their own feed.
Built only from forwarded posts we have actually read, on both sides. Coverage is early and deliberately incomplete: a missing link means we have not read the post that would prove it, never that the relationship does not exist. Counts are distinct forwarded posts observed, so they only ever go up as we read more.
Cite this entry
A live page changes as we take new readings, so a citation should name the measurement it is based on, not just the URL. The line below cites the subscriber count as measured 11 August 2026 — this entry's latest reading, not the date you are reading this.
“Agili” (@agili_channel), 32 subscribers as measured 11 August 2026. Telegram Register, tgregister.com/channel/agili_channel.
Full measurement history, CC BY 4.0. Every reading this register holds for this entry, not just the latest one, as a dated, downloadable record: CSV · JSON. Free to use with attribution to tgregister.com. Each file carries its own generation timestamp, which is the figure to cite for exactly when the data was retrieved.