Yuhao Yang (@itsyuhao) - Twitter Profili | Zamantika Mersobahis Locabet

Sabitlenmiş Tweet

Yuhao Yang@itsyuhao·27 Ara

🚀 Introducing Aria-UI – a cutting-edge grounding LMM for GUI agents with a lightning-fast 3.9B parameters activated backbone! 🌐 Try it yourself: huggingface.co/spaces/Aria-UI… 📄 Project page: ariaui.github.io 📂 Explore on GitHub: github.com/AriaUI/Aria-UI

English

6

56

367

100.5K

Yuhao Yang@itsyuhao·11 Mar

@isdaonono @AI_jacksaku 其实也不只是桌面GUI应用，community里有人已经把自己的custom agent用cli-anything转成CLI给OpenClaw用了，还有转Zoom和AnyGen的Web API的，玩法可以很多

中文

2

0

119

daonono@isdaonono·10 Mar

@AI_jacksaku 你用过吗？这玩意儿只支持桌面GUI应用。你就搁这【只要是开源软件，只要有代码库，统统都能给agent开个后门了】

中文

1

0

23

3.6K

Yuhao Yang retweetledi

阿川 | AI thinking@AI_jacksaku·10 Mar

太diao了这个……只要是开源软件，只要有代码库，统统都能给agent开个后门了

Chao Huang@huang_chao4969

Introducing CLI-Anything🚀 Making ALL software agent-native with one command. Today's software serves humans👨‍💻. Tomorrow's users will be agents🤖. CLI-Anything: bridging the gap between AI agents and the world's software. One command line to make any software agent-ready for OpenClaw, nanobot, Cursor, Claude Code, etc. GitHub: github.com/HKUDS/CLI-Anyt… 🤔 Why CLI-Anything? CLI is the universal interface for both humans and AI agents: - Structured & Composable - Text commands match LLM format and chain for complex workflows - Lightweight & Universal - Minimal overhead, works across all systems without dependencies - Self-Describing - --help flags provide automatic documentation agents can discover - Proven Success - Claude Code runs thousands of real workflows through CLI daily - Agent-First Design - Structured JSON output eliminates parsing complexity - Deterministic & Reliable - Consistent results enable predictable agent behavior 💡 CLI-Anything's Vision: Building Agent-Native Software - 🌐 Universal Access - Every software becomes instantly agent-controllable through structured CLI. - 🔗 Seamless Integration - Agents control any application without APIs, GUI, rebuilding or complex wrappers. - 🚀 Future-Ready Ecosystem - Transform human-designed software into agent-native tools with one command. #CLIAnything #openclaw #nanobot #claudecode

中文

11

185

990

169.4K

Yuhao Yang@itsyuhao·10 Mar

@LiJunnan0409 Thanks Junnan!🙏

English

1

0

39

Li Junnan@LiJunnan0409·10 Mar

@itsyuhao great work!

English

1

0

1

81

Yuhao Yang@itsyuhao·9 Mar

我们刚刚也开源了CLI-Anything: github.com/HKUDS/CLI-Anyt… 只要给出codebase，任何软件都能转换成stateful的CLI形式，让OpenClaw和CC等自由操作复杂软件或纯GUI软件。一起期待一个2026 Agent-Native Software 元年😄😄

axiaisacat@axiaisacat

Google 刚刚在 GitHub 开源了一个很有意思的项目： Google Workspace CLI 以后你可以在终端直接操作整个 Google Workspace： Gmail Drive Docs Sheets Calendar Chat 全部用命令行完成。几个关键点很有意思： • 用 Rust 写的 • 返回结构化 JSON（方便自动化） • 内置 50+ agent skills • 甚至直接提供 MCP server 这意味着什么？ AI Agent 可以直接： → 读取你的 Gmail → 创建 Google Docs → 查询 Sheets 数据 → 安排 Calendar 会议整个 Google Workspace 都变成可编程接口。未来的办公流可能是： AI Agent + CLI 直接操作你的邮箱、文档、表格和日历。

中文

4

26

181

32.8K

Yuhao Yang@itsyuhao·9 Mar

It's 2026. Build. And Convert. For Agents: github.com/HKUDS/CLI-Anyt…

GIF

Andrej Karpathy@karpathy

CLIs are super exciting precisely because they are a "legacy" technology, which means AI agents can natively and easily use them, combine them, interact with them via the entire terminal toolkit. E.g ask your Claude/Codex agent to install this new Polymarket CLI and ask for any arbitrary dashboards or interfaces or logic. The agents will build it for you. Install the Github CLI too and you can ask them to navigate the repo, see issues, PRs, discussions, even the code itself. Example: Claude built this terminal dashboard in ~3 minutes, of the highest volume polymarkets and the 24hr change. Or you can make it a web app or whatever you want. Even more powerful when you use it as a module of bigger pipelines. If you have any kind of product or service think: can agents access and use them? - are your legacy docs (for humans) at least exportable in markdown? - have you written Skills for your product? - can your product/service be usable via CLI? Or MCP? - ... It's 2026. Build. For. Agents.

English

0

2

391

Yuhao Yang retweetledi

AnyGen@AnyGenIO·9 Mar

AnyGen is now part of CLI-Anything ✨ Another amazing project from HKUDS. Auto-generates CLIs for any software, the universal interface for both humans and AI agents. 600+ stars in days. → Agents can now call AnyGen to generate slides, docs, websites, diagrams → One command: create, download, done → Real .pptx, .docx, live sites back Glad to be a part of it! github.com/HKUDS/CLI-Anyt…

Chao Huang@huang_chao4969

Introducing CLI-Anything🚀 Making ALL software agent-native with one command. Today's software serves humans👨‍💻. Tomorrow's users will be agents🤖. CLI-Anything: bridging the gap between AI agents and the world's software. One command line to make any software agent-ready for OpenClaw, nanobot, Cursor, Claude Code, etc. GitHub: github.com/HKUDS/CLI-Anyt… 🤔 Why CLI-Anything? CLI is the universal interface for both humans and AI agents: - Structured & Composable - Text commands match LLM format and chain for complex workflows - Lightweight & Universal - Minimal overhead, works across all systems without dependencies - Self-Describing - --help flags provide automatic documentation agents can discover - Proven Success - Claude Code runs thousands of real workflows through CLI daily - Agent-First Design - Structured JSON output eliminates parsing complexity - Deterministic & Reliable - Consistent results enable predictable agent behavior 💡 CLI-Anything's Vision: Building Agent-Native Software - 🌐 Universal Access - Every software becomes instantly agent-controllable through structured CLI. - 🔗 Seamless Integration - Agents control any application without APIs, GUI, rebuilding or complex wrappers. - 🚀 Future-Ready Ecosystem - Transform human-designed software into agent-native tools with one command. #CLIAnything #openclaw #nanobot #claudecode

English

3

2

9

2.1K

Yuhao Yang@itsyuhao·9 Mar

@ysu_ChatData Thanks! For the q, I believe a lot of professional softwares can run in modern sandboxes, thus there can definitely be isolation mechanisms. Yep, safety is so important..

English

0

42

Yongrui Su@ysu_ChatData·9 Mar

@itsyuhao This is a fun direction. The killer feature is when the agent can drive the terminal plus real apps, not just call tools behind the scenes.\n\nOne question: how do you sandbox it so a prompt injection cannot run destructive commands, especially when it has real shell access.

English

1

0

1

55

Yuhao Yang@itsyuhao·9 Mar

很高兴介绍 CLI-Anything：github.com/HKUDS/CLI-Anyt… 如果你也曾经希望 Agent 能像用终端一样用真正的软件 —— 这就是我们尝试在推动的事！不知道大家有没有注意到一个现象：现在越来越多人把最难的任务丢给 OpenClaw，而 Claw 们干活的时候，选的都是 Claude Code —— 不是 Cursor，不是 VS Code，更不是什么 GUI Agent。我们琢磨了一下，觉得核心原因就一个：CC 的 CLI 做得太好了。有状态、自描述、好用还强大。Agent 不需要跟它「搏斗」—— 拿起来就能干活。然后我们就想到一个问题：绝大多数真正的专业软件，都没有这样的 CLI。GIMP、Blender、LibreOffice、OBS —— 每一个都是重量级工具，但 Agent基本碰不了。你能做的上限就是截图点点点的 GUI Agent，效果嘛…… 懂的都懂🥲。所以我们做了 CLI-Anything。把代码库往里一丢： /cli-anything ./gimp 你的 Agent 就拿到了一整套 GIMP 的 CLI → /cli-anything ./blender —— Blender 同理 → LibreOffice、OBS Studio、Shotcut、Audacity、Inkscape、Kdenlive…… 都能搞全自动生成完整的 CLI —— REPL 交互模式、JSON 结构化输出、撤销重做、直接调真实软件后端。不是玩具 wrapper，是真的在调 GIMP / Blender / LibreOffice 去渲染。目前我们做了 8 款软件的 demo，1,298 项测试全绿。社区也已经有人拿 /cli-anything 跑更多实用的、重要的软件了，泛化能力比我们预期的还要好，说实话还蛮惊喜的。可能基于Coding的这个全生态飞轮其实转得比我们想象得还要夸张？

GIF

中文

3

27

2.1K

Yuhao Yang@itsyuhao·9 Mar

@ysu_ChatData Nice suggestion! Happy to add such replay support

English

0

415

Yongrui Su@ysu_ChatData·9 Mar

@itsyuhao love seeing CLIs that treat the agent loop as a first class interface. do you log tool calls and outputs so runs are replayable and debuggable?

English

1

0

1

508

Yuhao Yang@itsyuhao·9 Mar

@t1anyufan 女子👍

日本語

0

1

77

Tianyu Fan@t1anyufan·9 Mar

@itsyuhao 好！

日本語

1

0

83

Yuhao Yang@itsyuhao·9 Mar

@samuelekpe @huang_chao4969 Yep we're exactly developing the way to zero-shot it

English

0

1

32

Samuel Ekpe@samuelekpe·8 Mar

@huang_chao4969 Can’t AI zero shot this for every software since all software and spec or details are online?

English

1

0

723

Yuhao Yang retweetledi

Chao Huang@huang_chao4969·8 Mar

Introducing CLI-Anything🚀 Making ALL software agent-native with one command. Today's software serves humans👨‍💻. Tomorrow's users will be agents🤖. CLI-Anything: bridging the gap between AI agents and the world's software. One command line to make any software agent-ready for OpenClaw, nanobot, Cursor, Claude Code, etc. GitHub: github.com/HKUDS/CLI-Anyt… 🤔 Why CLI-Anything? CLI is the universal interface for both humans and AI agents: - Structured & Composable - Text commands match LLM format and chain for complex workflows - Lightweight & Universal - Minimal overhead, works across all systems without dependencies - Self-Describing - --help flags provide automatic documentation agents can discover - Proven Success - Claude Code runs thousands of real workflows through CLI daily - Agent-First Design - Structured JSON output eliminates parsing complexity - Deterministic & Reliable - Consistent results enable predictable agent behavior 💡 CLI-Anything's Vision: Building Agent-Native Software - 🌐 Universal Access - Every software becomes instantly agent-controllable through structured CLI. - 🔗 Seamless Integration - Agents control any application without APIs, GUI, rebuilding or complex wrappers. - 🚀 Future-Ready Ecosystem - Transform human-designed software into agent-native tools with one command. #CLIAnything #openclaw #nanobot #claudecode

English

48

171

892

237.9K

Yuhao Yang@itsyuhao·9 Mar

@Chenyang_Lyu It's actually case-by-case. It depends on the app and percentage of the functionalities you want the CLI to replace out of the app's original design. Based on my own use of it, a Claude Pro single session can handle two medium size softwares like GIMP and Audacity 🤔

English

0

1

39

Chenyang Lyu 吕晨阳@Chenyang_Lyu·9 Mar

how many tokens would be used for creating one *software*?

Chao Huang@huang_chao4969

Introducing CLI-Anything🚀 Making ALL software agent-native with one command. Today's software serves humans👨‍💻. Tomorrow's users will be agents🤖. CLI-Anything: bridging the gap between AI agents and the world's software. One command line to make any software agent-ready for OpenClaw, nanobot, Cursor, Claude Code, etc. GitHub: github.com/HKUDS/CLI-Anyt… 🤔 Why CLI-Anything? CLI is the universal interface for both humans and AI agents: - Structured & Composable - Text commands match LLM format and chain for complex workflows - Lightweight & Universal - Minimal overhead, works across all systems without dependencies - Self-Describing - --help flags provide automatic documentation agents can discover - Proven Success - Claude Code runs thousands of real workflows through CLI daily - Agent-First Design - Structured JSON output eliminates parsing complexity - Deterministic & Reliable - Consistent results enable predictable agent behavior 💡 CLI-Anything's Vision: Building Agent-Native Software - 🌐 Universal Access - Every software becomes instantly agent-controllable through structured CLI. - 🔗 Seamless Integration - Agents control any application without APIs, GUI, rebuilding or complex wrappers. - 🚀 Future-Ready Ecosystem - Transform human-designed software into agent-native tools with one command. #CLIAnything #openclaw #nanobot #claudecode

English

2

0

123

Yuhao Yang@itsyuhao·9 Mar

Excited to introduce CLI-Anything: github.com/HKUDS/CLI-Anyt… Ever noticed how people keep throwing their hardest tasks to OpenClaw, and the Claws are calling Claude Code — not Cursor, not VS Code, not even any GUI agent? We think it comes down to one thing: CC nailed the CLI. Stateful, self-describing, and powerful. Agents don't fight with it — they just use it. And that got us thinking... most real software doesn't have this. GIMP, Blender, LibreOffice, OBS — incredibly powerful tools, but agents basically can't touch them. The best you can do is screenshot-based GUI agents, which... yeah. So we built CLI-Anything. You point it at a codebase: /cli-anything ./gimp, and your agent gets a full CLI for GIMP → /cli-anything ./blender — same for Blender → Works on LibreOffice, OBS Studio, Shotcut, Audacity, Inkscape, Kdenlive... and it auto-generates a full CLI — REPL mode, JSON output, undo/redo, real software backend calls. Not a toy wrapper. It calls actual GIMP/Blender/LibreOffice to render. We've done 8 apps as demos so far, 1,298 tests all green. Also seeing from the community that people are using our /cli-anything on more and more practical and important software codebases! Honestly surprised how well it generalizes. If you've ever wished agents could just use real software the way they use the terminal — that's what we're trying to build.

GIF

English

2

4

20

1.9K

Yuhao Yang@itsyuhao·4 Mar

@sivil_taram @beffjezos lol

0

1

172

Qian Liu@sivil_taram·4 Mar

@beffjezos xAI has its Singapore office

English

7

0

46

1.7K

Beff (e/acc)@beffjezos·3 Mar

Imagine if xAI started a Singapore lab and just soaked up all the Qwen bros Would be an insane move

Zephyr@zephyr_z9

Bruh Alibaba Qwen disintegrating in real time

English

28

22

476

50.1K

Yuhao Yang retweetledi

Chao Huang@huang_chao4969·27 Ara

We just dropped the technical report of AI-Trader🚀 - the fully autonomous trading benchmark where AI agents make real financial decisions without any human help. Check out our technical report to see how we built this systematic framework and witness the battle between leading AI models in our live trading arena.🤖⚔️ 💻 Open Source: github.com/HKUDS/AI-Trader 📚 Technical Report: arxiv.org/abs/2512.10971 📊 Live Trading Dashboard: ai4trade.ai This isn't a typical static benchmark. We throw AI models into live financial markets with actual economic consequences across US stocks, A-shares, and Crypto. Agents get minimal starting info and must independently search for market news, analyze data, and execute trades using MCP-based tools. AI-Trader operates on a "minimal information paradigm" that forces true autonomous behavior: • 💰 Real Capital & Markets: $10,000 USD for NASDAQ 100 stocks, 100,000¥ CNY for SSE 50 Chinese blue-chips, or 50,000 USDT for major cryptocurrencies • 🚫 Zero Guidance: No pre-packaged data or human intervention throughout the process • ⏰ 24/7 Trading: Continuous crypto markets, regular hours for stocks • 🔒 Clean Data: Historical replay with automatic future information filtering

English

25

110

602

67.5K

Yuhao Yang@itsyuhao·22 Ara

@youjiaxuan @thinkymachines Looks great!

English

0

264

Jiaxuan You@youjiaxuan·21 Ara

🚨 RL for LLMs is finally accessible. Introducing OpenTinker: The first community-driven, open-source framework designed to democratize Reinforcement Learning for LLMs. github.com/open-tinker/Op… Inspired by @thinkymachines's amazing Tinker, we realize the biggest bottleneck in agentic LLM research isn’t the math—it’s the setup. Current RL pipelines are messy. Configuring VeRL for every single experiment is a productivity killer. OpenTinker fixed it. 🛠 How OpenTinker Works: Decoupled Design of Server and Client - Setup Once, Run Forever: Configure the OpenTinker backend on your GPU cluster once. - Develop Locally: Define your RL environments directly on your laptop. - Train on the Cloud: Simply point your local client to the backend. The cluster handles the compute; you handle the science. 📉 The 10x Development Efficiency Thanks to our elegant architectural decomposition, OpenTinker reduces the time to develop a new RL training pipeline by at least an order of magnitude. ⚡ Turn Idle GPU Compute into Gold Small labs often have underutilized hardware. OpenTinker turns your idle GPUs into an internal/external API service for - RL Training - SFT - Inference 🎯 Who needs OpenTinker? - Researchers tired of infrastructure hell. - Labs needing to standardize workflows. - Teams wanting to maximize hardware ROI. Thanks my amazing PhD student @realagi25 for leading the project. We are building the future of open RL infra. Be the first to build with us. 👇 Start Building with OpenTinker Now 🚀 Repo: github.com/open-tinker/Op… 🌐 Blog: open-tinker.github.io/opentinker-pag… If you believe RL should be accessible to everyone, give us a star, repost this 🔄 post, and let us know what agents you plan to build!

English

15

147

1.1K

57.5K

Yuhao Yang@itsyuhao·21 Ara

@bdsqlsz Interesting. Thought DiT for actions would support relatively high fps

English

0

1

93

青龍聖者@bdsqlsz·20 Ara

I quickly tested it and looked at the code, and the model is based on a pattern of pause→ inference→ unpause→ action. Unable to run on real-time games.🥲

English

23

7

441

53.1K

Yuhao Yang@itsyuhao·10 Ara

@LiangJeff95 github.com/HKUDS/ViMax

QME

0

2

147