My massive article on Roblox Optimization is out! How massive? It has chapters!
It comes with an uncopylocked game called Cardboard Box Simulator, which has been optimized to run at 60fps on low spec phones at full quality. It also contains a whole bag full of neat tricks and approaches for making your own builds run well.
https://t.co/MFHsp4MUlj
#Roblox #robloxdev
So GRI had a gameplay problem: Some characters have wallhacks as a skill. This made trees you could shoot through way too effective. The solution: make the trees solid. This had its own challenges, but now we have fuzzy trees! #RobloxDev
@Khaelises Yup! The trees were procedurally generated (SDF blobs) and the cards are bands that run right the way around the mesh in a grid shape with a carefully made tiling texture. No normal manipulation - just transparency to get rid of the SSAO.
@TrenchWarsRBX Yeah the whole game is set up to be fully destructible now, so there’s nowhere to hide but there’s no wall bangs either. If you can shoot someone they can see you and shoot you too.
Good ol' building destruction in Roblox. One of my coworkers said my buildings looked like the toolbox default happy house and I gotta say I took that kinda personally. #RobloxDev
I have a "spore creature creator" going in Roblox! Any leg count and body shape works. And there is zero chance I'm going let any of you degenerates play with it because I know what you'll make... #robloxdev
🌈 Color3 just got three new methods: GetContrastingColor, GetContrastRatio, and GetLuminance 💡
https://t.co/fz9XOGpsZ7
See how good you are at guessing color contrast in the demo game and be humbled by how bad you are at guessing luminance!
https://t.co/eml7vM0b6m
I've been messing around a lot with EditableImages and CaptureService a lot lately and wrote up a cute tutorial on how you can use them together to create cool in-game effects!
https://t.co/BLfhqCt5vf
#roblox#robloxdev
Welp. Thats it - this is what alien intelligence looks like, folks. Completely orthagonal to something humans are equipped to do or even really comprehend.
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.
We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵
Blog: https://t.co/bOWCcFhCVq