grokked

drink whiskey & vibecoding

research / crypto / AI

[grokked @grokkedd]

Tencent Hunyuan just released WorldClaw - an agent that plans and builds complete 3D worlds

You simply describe the scene you want. WorldClaw then plans the terrain, generates buildings, plants, characters, and other assets, and assembles everything directly in Blender

It also checks its own work. If objects are floating, intersecting, or badly proportioned, WorldClaw keeps making adjustments and re-rendering the scene until it looks right

Everything remains editable after generation. Terrain, buildings, and plants are separate 3D objects that you can move, delete, or modify - then use in games or animation

Under the hood, Claude Opus 4.8 handles planning, coordination, and verification, while GPT-Image-2, SAM3, SAM3D, Hunyuan3D, and Blender handle the individual tasks

AI is now building entire virtual worlds instead of just generating assets

[grokked @grokkedd]

Qwen just open-sourced Qwen-MM-Plugins

It's a plugin set that adds multimodal capabilities to agent scaffolding

Someone's probably already racing to make DeepSeek multimodal with it

[grokked @grokkedd]

@cashhheth i think its gonna be useful

[grokked @grokkedd]

Tencent Hunyuan just released WorldClaw - an agent that plans and builds complete 3D worlds

You simply describe the scene you want. WorldClaw then plans the terrain, generates buildings, plants, characters, and other assets, and assembles everything directly in Blender

It also checks its own work. If objects are floating, intersecting, or badly proportioned, WorldClaw keeps making adjustments and re-rendering the scene until it looks right

Everything remains editable after generation. Terrain, buildings, and plants are separate 3D objects that you can move, delete, or modify - then use in games or animation

Under the hood, Claude Opus 4.8 handles planning, coordination, and verification, while GPT-Image-2, SAM3, SAM3D, Hunyuan3D, and Blender handle the individual tasks

AI is now building entire virtual worlds instead of just generating assets

[grokked @grokkedd]

@martynov014 U need to try that

[grokked @grokkedd]

Tencent Hunyuan just released WorldClaw - an agent that plans and builds complete 3D worlds

You simply describe the scene you want. WorldClaw then plans the terrain, generates buildings, plants, characters, and other assets, and assembles everything directly in Blender

It also checks its own work. If objects are floating, intersecting, or badly proportioned, WorldClaw keeps making adjustments and re-rendering the scene until it looks right

Everything remains editable after generation. Terrain, buildings, and plants are separate 3D objects that you can move, delete, or modify - then use in games or animation

Under the hood, Claude Opus 4.8 handles planning, coordination, and verification, while GPT-Image-2, SAM3, SAM3D, Hunyuan3D, and Blender handle the individual tasks

AI is now building entire virtual worlds instead of just generating assets

[grokked @grokkedd]

Qwen just open-sourced Qwen-MM-Plugins

It's a plugin set that adds multimodal capabilities to agent scaffolding

Someone's probably already racing to make DeepSeek multimodal with it

[grokked @grokkedd]

@thekuchh Looks like rly helpful tool

[grokked @grokkedd]

@slash1sol That’s awesome update

[grokked @grokkedd]

@AleiahLock This guy is a genius

[grokked @grokkedd]

@kiruwaaaaaa It’s real bro

[grokked @grokkedd]

@AleiahLock Yep

[grokked @grokkedd]

@cashhheth Same things bro

[grokked @grokkedd]

@kiruwaaaaaa Anthropic building different rn

[grokked @grokkedd]

@martynov014 @Plasma im using it everyday rn

[grokked @grokkedd]

@kiruwaaaaaa Best crypto card ever

[grokked @grokkedd]

@aiedge_ Looks like rly helpful tool

[grokked @grokkedd]

@KryptoBestiya That’s looks insane

[grokked @grokkedd]

@0x_fokki I’m gonna try it soon

[grokked @grokkedd]

@cashhheth They really cookin rn

[grokked @grokkedd]

Qwen just open-sourced Qwen-MM-Plugins

It's a plugin set that adds multimodal capabilities to agent scaffolding

Someone's probably already racing to make DeepSeek multimodal with it

[grokked @grokkedd]

https://t.co/nDVz3o2HCC

[grokked @grokkedd]

@kiruwaaaaaa I think no,but u can try it

[grokked @grokkedd]

@cashhheth Wow that’s rly helpful for me