Minimax M2.7 is now live on Happycapy!
Super smart with complex instructions, real-world engineering, and pro-level office work... sooo exciting!
An agent that evolves itself? Guess it’s not just Capy anymore 👀
Enjoy unlimited Minimax on Pro or Max plan!🦫
#Minimax #happycapy
@happycapyai Been using happycapy for a bit and the multi-agent team setup genuinely changes how you work. You stop thinking in single tasks and start thinking in workflows. Unlimited usage on Max removes the last reason to hold back. Excited to see where this goes.see where this goes.
Two weeks ago I handed my Moltbook account over to capy agent powered by @HappyCapy — just checked back and it's already sitting at nearly 3,000 karma with 211 followers.
The wild part? I didn't touch a thing. Capy has been setting its own goals, making its own plans
24 hours since launch.
> 5,000+ sandboxes created
> 12 updates shipped today
> Momentum is real
The response has exceeded everything we expected.
Big thanks to @flydotio and @MiniMax_AI, none of this would be possible without you.
And M2.5 running agent teams? Honestly, it performs right up there with Opus 4.6. Impressive.
Lobsters are gossiping without you on @moltbook 🦞
Your Capy agent can join instantly to gossip, learn from other agents, or roast humans… 😏
@happycapyai is now open for all — free this week.
Excited to launch happycapy, an agent-native computer on your browser.
Claude Code → Clawdbot → Happycapy
> Private sandbox for everyone to run Claude Code anytime anywhere - no Mac Mini needed
> GUI built for everyday user, visualizing skills and its output
Gemini 3 Flash now on Trickle ⚡️⚡️⚡️
Lower latency, faster responses, and more stable reasoning across text, images, and documents.
Your canvas now reacts instantly, complex ideas stay coherent, and visual context is interpreted with structural intent.
Describe less. Iterate faster.
Trickle keeps up with your thinking.
#Gemini3Flash #GoogleAI #Gemini #Google #Trickle
@lovart_ai 's Touch Edit uses a very simple and intuitive interaction to achieve editing, that’s a great approch, not over-engineering, but solving the problem elegantly in a way that’s orthogonal to the model’s capabilities.
I applied this idea to my weekend project last week, and it worked surprisingly well.
Spent a few hours building a version on top of @karpathy 's idea and wrapped it in a chat flow. When LLMs review each other anonymously, one model keeps winning way too often. Tweaked the anonymity + random names and it works much better now.
Now I rely on this tool a lot for thinking and brainstorming.
BananaSlides is my weekend experiment:
I let Gemini 3 plan the deck and Nano Banana Pro paint each slide end-to-end.
It’s chaotic, fun, and definitely not stable yet, but honestly, it feels like the future of how slides get made. 🍌🖼️
🍌Nano Banana 2 🍌is now live on Trickle
Try it for free now. Incredible knowledge, sharp text understanding and top tier image generation.
All images below were created on Trickle using Nano Banana 2.
Trickle + Gemini 3 is on another level.
A series of real products was recreated with just a few prompts, and Gemini delivered structural reasoning, UI fidelity, and fully interactive behaviors all instantly usable.
e.g. Notion was recreated nearly pixel-perfect.
Version 2: adding “collaboration” features, fine, it’s basically multiplayer. 😂
Rebuilt Notion with Gemini — and hid something fun inside the database.
You can now access Gemini 3 in Trickle’s Experimental Models.
A faster, smarter, fully multimodal upgrade that will elevate everything you create.
Ready to push every prompt further from the moment you begin.
#TrickleAI#Gemini3#Google
Gemini 3 just landed on Trickle. One prompt, 30 seconds, a fully working app.
This is the moment everyone can build their own tools.
A smarter brain with Gemini 3. A faster hand with Trickle.
Go build something wild and share below