Timothée Peter

Live jacobian computation through a function that contains an iterative solver. Hope to release this soon.

@pirodir This demo is Mk14. I'm currently building Mk16. I wrote them all, yes but with the help of llms for the latest ones.

@TimotheePeter How long did it take for dekal to reach this level of functionality? It’s written by you alone? I understand you are demoing a feature but CAD as REPL in general is so cool!

Here's me modeling a spherical five bar linkage using code

@__Kinya @JacklouisP pretty much where we're headed indeed

@JacklouisP @TimotheePeter Differentiable simulation engines are the next frontier for mechanical design. Being able to describe kinematics in code while leveraging gradient-based optimization for workspace mapping is a huge efficiency gain over traditional CAD iterations. 🤖📐 #Kinematics #Robotics

Just discovered Dekal via @TimotheePeter's video modeling a spherical five-bar linkage in code. What makes it interesting: > Code-based kinematic design with a differentiable simulation engine > Describe mechanisms in code, instantly get workspace mapping, reachability analysis, and static load testing across every joint > Idea to kinematic validation ASAP; then push to CAD Its a cool project built by a mechanical engineer frustrated with iteration speed in existing tools.

@basikthings Thanks! Current plan is to monetize it, yes. Not against having a free license for non-commercial use though.

@TimotheePeter Hey @TimotheePeter, super interesting! Do you plan to monetize this somehow or will you be open sourcing this?

If you want to try it yourself: demo.dekal.sh

demo.dekal.sh Try a live demo of Dekal's code based geometric constraint solver for CAD and robotics kinematics design.

A spherical five bar linkage, modeled using dekal.sh

Functional programing is a good fit for an implicit kernel, since modeling operations are functions themselves. Perhaps logic programing is a good fit for explicit (brep) kernels since the modeling operations are database mutations that might fail?

We are extending this with a static force and torque solver

Testing the compiler's ability to solve kinematics in real time. This also lets you extract positions or transmission ratios for analysis.

@beffjezos @elodinsystems seems to have a Rust port - news.ycombinator.com/item?id=396248…

@beffjezos Are you all writing your libraries in python? Would be cool to have access to Jax from OCaml or Rust

Extropic libraries for Thermo ML are built in JAX. Open sourcing coming soon. Long JAX.

Building a tool that lets you describe a mechanical assembly kinematic structure, parametrize it and visualize it in multiple assembly configurations.

@OnurGumusDev Approximating with a rectangular bounding box is perfectly fine to start off with. I'll have a look, thanks!

Cool idea! Right now, the packer only supports rectangular boxes (rectangular prisms), so it’s best suited for packing items that can be approximated with bounding boxes. If your 3D printer parts are roughly box-shaped, you could definitely use it to arrange them efficiently on a print bed by treating the bed as the container and setting the Z-axis height to the part height. That said, it doesn’t yet support arbitrary mesh shapes like STL files, so it won’t account for curves or voids in the model.

🚀 Just added a new feature to my F#-powered 3D bin packer: You can now drag and drop boxes after the algorithm runs perfect for fine-tuning your Tetris skills 🧠📦 Try it out: 3dpack.ing

Cool work by @networked_af with the airlab. Check it out on HN: news.ycombinator.com/item?id=441893…

@simonpj0 @TimSweeneyEpic would be cool to be able to use Verse outside of the Unreal Editor for Fortnite! Any plans/timeline for releasing Verse as a standalone language and compiler?

@ThePrimeagen, thoughts on F# compared to ocaml? F# seems to get me off the ground a lot faster than ocaml