Vibe Coding a AAA Racing Game with ThreeJS day 5 6
✅ 1K textures (up from 256)
✅ Better Tunnels
✅ Motion Blur
✅ Tire Smoke
✅ Environment reflections on the car
✅ Huge improvements to the road building systems (different road types, decals, and skinned props)
All props and textures generated with codex
🚀 Introducing Articraft, a coding agent for articulated 3D asset creation.
Articraft writes code, executes it, receives validation feedback, and refines the result into simulation-ready 3D assets with parts, joints, and motion.
We’re also releasing Articraft-10K: 10,000 articulated objects across 250 categories, unlocking large-scale interactive scenes for robotics simulation and physical AI.
🔗 Project page: articraft3d.github.io/
💻 Code: github.com/mattzh72/articraf…
Vibe Coding Metahuman with ThreeJS: Hairs and Combs
It's now possible to draw hair, density and flow masks, place guide curves and comb them!
The hair is rendered partially as curves and card meshes to achieve more volume with less strands and there are lots of material/shader settings to tweak it afterwards. Don't ask me what they all do, the melanin slider for sure does something
Okay it's not photo realistic but i'd say pretty solid for a day's work?
The UI is still a complete mess but i'll clean it all up eventually
That's it for the head! Next up is tedious modeling work again, on to the body!
vibe-stack.github.io/vibe-hu…
Vibe Coding Metahuman With ThreeJS and Codex: Skinning the model
So I think gpt image 2 a good upscaler can reliably produce different ethnicities and skin textures
Both skins were created in a single prompt in chatgpt, so i can probably plug @fal to the repo and integrate it into the UI, so everyone can create different skins with a prompt
I think this concludes the head part! Next up i'll try to slightly tap into hairs, though i'm worried they might be more complex than the whole head!
After that moving on to the body
oh and i've added draggable control points directly on the mesh so it's easier to sculpt the face instead of knowing where the hell the lower nose cartilage is on a slider
Demo:
vibe-stack.github.io/vibe-hu…
the future is a complete , algorithm driven natural blend of visual splats and physical meshes. Meshes needs to be physically consequential to stay relevant
...can someone create a wasm core web3dengine that does rendering and physics in ONE world simulation.. (physics can opt-out without dragging rendering performance)
Qwen3.5 27B is awesome (the entire family above 9B is impressive). You can now try it directly in your browser at SOTA speeds with whatever GPU you have: hf.co/spaces/Ex0bit/tensorbe…
My previous research in practice - The `Intel/Qwen3.5-27B-int4-AutoRound` is particularly good.
A 27B model is #2 on pinch-bench
You’d need 150,000$ in GPU hours to train this from scratch (base post training)
Basically 1-2 weeks over 256 H100s
That is not unreasonable, you’d need 540B tokens for pre-training and a bit more for post training.
None of this is crazy
The future of webAI webGPU webNN would be amazing . Building an product that taps into the full potential of these would be the wise thing to do for any digital nomads out there
used many coding AI for the past 1.5 years to create my 3d engine and online sandbox - the project is now at 400,000 lines of code spanning across frontend, backend, 3d engine core, etc. My personal take: Claude 4.5 thinking, codex high are tier1, everything else no even close