Soooo I have added automated agentic CAD reviews between any two agents (1 MCP + 1 API)
And agents can create RL strategies, run them on cloud GPU through @modal, then reflect on them and improve
Also working on adding automated finite element analysis with feedback loop
If anyone wants to try, send me DM
Yesterday @bubbleboi reposted my turbine comment, so naturally I expected to wake up to all his 110.4K followers following me too.
On a more serious note; how do reposts work, because they don鈥檛 show up on the reposters feed?
So... I'm already considernig throwing away this new design. Abduction first makes its range very limited
Should I swap it so I do hip pitch first, then abduction and then add another pair of motors for hip yaw?!
That would make the most versatile setup and I'm waiting for PCB anyway...
Hear me out;
LLMs are almost at the point where they don't need to become smarter for 99% of people (e.g. robotic dev)
Then it will be all about speed and I'm building a fully automated feedback loop app for robotics development
When LLMs start doing 1000+ tok/s, tool calls will become an INSANE bottleneck in comparison
So everything around the LLM needs to be FAST. No excuses














