Seven days ago, we started an experiment in Recursive Self-Improvement (RSI).
We wanted to explore a simple question: What happens when we let an AI agent iteratively improve its own harness?
So we let Lobe Agent run for an entire week, using our RSI framework to explore and
LobeHub 2.2.20 is here. And things just got smoother. ⚡
Your topics and docs now show up before the network even catches up. Yep, that fast.
And there's more:
🎬 Seedance 2.5 + Claude Haiku 5.5 just joined the party
🌿 Love a reply? Fork it into a whole new topic
📁 Your
Your Agent can generate videos now?? 🎬
Yep. Video generation just landed in LobeHub.
Tell your Agent what you want to see, and it'll make it happen.
Text to video. Multiple models. All in one workspace.
No switching apps. No extra steps.
Go give it a spin:
Okay, this was a fun one. 🔥
We put Muse Wonder and Gemini 4 Argon head-to-head to see how well they could build a Terraria-style game from scratch.
Same prompt. Same thinking effort. No excuses.
And wow, BOTH models were seriously fast.
But Gemini 4 Argon? It absolutely
🚨 Google is back in the AI race. And it came swinging.
Gemini 4 Argon is here. Now live on LobeHub.
And this thing means business.
🧠 Serious reasoning power
💻 Built for complex coding and agent workflows
🏆 #1 on the Vals Index
⚡ 1M-token context
Another heavyweight just
