Sydney, New South Wales
Joined October 2012
Been looking for a new browser home since @arcinternet announced its sunset. Ardia by @barre_of_lube feels closest to Arc so far.
Introducing Ardia, the open-source Arc I've been working on this for a few months now. It's feature-complete, already used by a bunch of former Arc users, and extremely performant. github.com/mantrakp04/ardia
2
21
1,384
GPT-6 made computer use the frontier. We make it open. 🚀 Meet OpenGCUA, an ongoing project to open the full stack for generalist computer use. All 8 frontier models improve on OSWorld-v2, GPT-6 Astra included. On average: +19.6 points and 53% fewer steps than batched GUI-only. The idea: let the model decide how to observe, act and organize its work, across visual and programmatic interfaces. We call such a system a generalist computer-use agent (GCUA). Two designs make it work: ⚙️ Code execution: code becomes the model's only action space. Each step is one program that can combine GUI actions, shell commands and app APIs, with loops and checks 🧭 Harness-to-model decision transfer: As models become more capable, we envision progressively shifting selected workflow decisions from fixed harness rules to the model Open today: 📄 Tech report 🧠 OpenGCUA harness and prompts, with an OSWorld integration example 📜 Evaluation trajectories of all 8 models 💻 macOS real-device harness, as a native menu-bar app (preliminary) 🔌 Cross-Device SDK (preliminary) 🧪 7 interactive labs and real-Mac demos Paper: arxiv.org/abs/2610.22200 Code: github.com/Tongyi-MAI/OpenGC… Labs and demos: tongyi-mai.github.io/OpenGCU… A closer look below 🧵
5
3
2
9
551
🚀 Meet MAI-Gym: an open platform for GUI agent evaluation & RL across mobile, desktop and web. 20,000 parallel environments, automated app/task synthesis, and open code, data & weights. 🔥 Our 9B agent scores 78.4% on MobileWorld and 76.2% on OSWorld-Verified.
2
6
209
Quyu Kong retweeted
我们谜底黑胶的设计上了 OpenAI 官网,成了别人的 Showcase,甚至连我们都没提一句 Our design of app MD Vinyl made it onto OpenAI’s website as someone else’s showcase, without even a mention of us. 😅 😅 @OpenAI @sama
21
19
9
277
61,615
Quyu Kong retweeted
1/ Introducing Cua Perception: Cua Driver can now see. Nothing to click? It reads the pixels. Canvases, games, remote desktops and custom-drawn apps become labelled regions an agent can act on, parsed locally on the CPU. github.com/trycua/cua
53
124
23
1,651
104,751
Quyu Kong retweeted
1/ Today we're announcing the stable Cua Driver release for Omarchy - a new foundation for computer use, built into the OS from the ground up. Over the last month, we worked directly with @dhh, @SpencerGBull and @vaxryy to bring a native synthetic cursor to Omarchy's Hyprland compositor, enabling true multi-cursor computer use at the OS level. Cua Driver is open source and available at github.com/trycua/cua
118
168
57
1,964
341,783
Jev @typesafeai gui agent demos from @trycua, @nielsmdt99 and others look insanely fast.🔥 But does that speed translate into reliable task completion on mobile-use? We tested the pair on MobileWorld’s 117 GUI-only tasks where Qwen3.8-Max plans, Jev acts (task demos below). 🧵
5
5
1
14
298
MobileWorld requires reasoning not just clicking. 🧠 Qwen3.8-Max reasons and sets short-horizon subgoals with visible completion criteria. ⚡ Jev selects and executes UI actions, then hands back for replanning. The planner thinks ahead. The actor handles the UI.
1
132
389 ms median Jev API latency. 27.4% task success. Qwen3.8-Max end-to-end reference: 77.8% (@Alibaba_Qwen). Fast decisions ≠ reliable completion. That’s why benchmarks matter beyond demos. Integration + full trajectories coming soon at github.com/Tongyi-MAI/Mobile….
2
93
🚀 Excited to share our latest work on foundation GUI agents! We present Qwen-UI-Agent, a real-world centric foundation GUI agent spanning mobile, computer-use, web, and DeepSearch environments. 🌐 Project Page: tongyi-mai.github.io/Qwen-UI… #LLM #Agent #GUI #tech
3
8
33
2,569
Introducing Qwen-UI-Agent — a real-world GUI agent for phones, computers, and browsers, with DeepResearch capabilities. Optimized for real-world environments, it combines GUI, CLI, and batched actions to complete complex, multi-step workflows—while retaining strong general reasoning and broader agentic capabilities. Qwen-UI-Agent achieves state-of-the-art performance on mobile and frontier-level performance across desktop and web. 🔗 Tech Report & Demos: tongyi-mai.github.io/Qwen-UI… Just check out our really cool demos and see how powerful it is! 😃
1
1
49
📣 MobileWorld benchmark updates: open benchmark for autonomous mobile agents — 201 tasks, 20 Android apps. ACL 2026 📄 New on the leaderboard: @AnthropicAI Opus 4.7 and @Kimi_Moonshot K2.6 — and they fixed the same bug. #LLM #AIagents 🐙 github.com/Tongyi-MAI/Mobile… 🧵👇
1
1
5
93
What still breaks SOTA models: 🛑 Mastodon Android↔web feature gaps 🛑 Long Mattermost threads blowing the step budget 🛑 PDF tasks stuck in search-and-scroll loops 🛑 Archive flows where date filtering eats every step Plenty of headroom.
1
27
🆚 New: head-to-head Arena — compare any two models side-by-side with screenshots + thinking traces. Community submissions open. 📝 tongyi-mai.github.io/MobileW… 🏆 tongyi-mai.github.io/MobileW…
20
Quyu Kong retweeted
MAI-UI is officially live! 🚀 We’re releasing a family of foundation GUI agents (2B to 235B) setting new SOTA benchmarks in GUI grounding & mobile navigation. 🏆 #1 on ScreenSpot-Pro (73.5%) 🏆 #1 on AndroidWorld (76.7%) 🔗 tongyi-mai.github.io/MAI-UI-…
1
1
1
127