It’s Brains season. Not the Halloween delicacy, but the accelerator. Have a big idea that benefits the public? Consider philanthropic funding. This program will introduce to the major players and help you reframe your pitch. spec.tech/brains
Blake Courter retweeted
We built the first GPU-accelerated metal casting solver.
FastCast by Digital Metal does fill, solidification, porosity & distortion on a single NVIDIA GPU, 91× faster than an industry-leading $15k/yr solver with the same level of accuracy.
Verified against real world results
Need some of these Claude session to PR tonight or it will be a mess tomorrow... Computer is getting hot from testing....
While Luke at GCL and @connorkapoor have made real parts with #OmegaAI, I'm a bit late to the game. My daughter sketched up this pattern for a floor register. We realized the space it's going is not quite square. Prompted Omega, refined the design, and sent to @sendcutsend ...
Blake Courter retweeted
Demo’ing this with our new notebook api monday at ntop, rsvp here: partiful.com/e/EYm3nB8M2MyNt…
When MCP with actual design tools for nTop, @brad_rothenberg
Implicit geometry is already basically math pretending to be CAD.
Feels like this should be THE design environment for an AI agent.
Give it the blocks. Let it cook.
Photo from a friend.
>>> Let's make the left model in this image.
Opus 5.5 and Astra (lighter maple):
#OmegaAI
Which woods can you recognize? Loosely based on mokumeproject.github.io/, with additional "research" for aging, heartwood / sapwood ratios and color, staining, etc. Next is spalting and worm holes...
Blake Courter retweeted
hill i’m willing to die on: parametric modeling is a scam. it doesn’t really work. it is slow. it is not a skill issue
Making code-based representations of CAD? Don't forget the first and second laws of feature interoperability:
1. You can make a scripting language that makes feature models in any CAD system.
2. Your scripting language cannot satisfactorily translate features between CAD systems.

















