Would anyone be interested in scaling completely different Agent environments? 👀
Teaser: We're open-sourcing AutoEnv — generate a full env (text, multimodal, even 3D) + scripts to continuously scale in-world data for just ~$4 💸
Textual scaling is available at: https://t.co/FPJvtyPE7B
See Generated Multimodal Env Below:
Video Reasoning is here.
And it’s not coming — it already beat everyone.
Sora-2, Veo-3.1-pro, Kling, MiniMax, Seedance…
All get 0% Exact Match on 3D mazes.
Does this mean “video reasoning” is a meme?
No.
It means nobody taught them how to reason yet.
1/6
Is ReAct the final for LLM Agents? 🤔
Not anymore. ReAct is stuck step-by-step. Planning agents separate plans from actions.
Introducing ReCode: agents control their decision granularity, like humans do.
20.9% better, 78.9% cheaper, 3.7× data efficient. The ReAct era is over.
1/5
🤖Check The Hitchhiker’s Guide to Agents HERE🤖
Our Foundation Agents Survey V2 level up to 396 pages – every chapter is a full-on survey itself!
🧠 Agent Framework & Components
🌍 World Model & Memory
🔄 Self-Evolution
👥 Multi Agents
🛡️ Safety
1/4
🤖Check The Hitchhiker’s Guide to Agents HERE🤖
Our Foundation Agents Survey V2 level up to 396 pages – every chapter is a full-on survey itself!
🧠 Agent Framework & Components
🌍 World Model & Memory
🔄 Self-Evolution
👥 Multi Agents
🛡️ Safety
1/4
🔥Excited to share our latest work: Agent KB
It achieves new open-source SOTA on the GAIA benchmark!
We enable agents to learn from each other's experiences across tasks through hierarchical experience sharing.
Paper📜: https://t.co/uCC4A4hb86
Code🧑💻: https://t.co/6E6m73Mrq2
It's actually a pity that we got no enough time to maintain OpenManus during the past 3 months.
But the better news is that we will build a formal open-source community for OpenManus at the end of this month.