AI is now improving its own model architectures and chips. The power infrastructure it runs on deserves the same attention.
We used RL to teach a Kimi base model to design medium-power transformers. It met 93% of unseen specs, helping to compress multi-week engineering efforts into a few minutes of inference.
1/5
Beyond new designs, we want to extend this to modular remanufacturing: keeping the core and tank from an existing transformer, then redesigning the windings, insulation and cooling for a new spec. That forces the model to design around components that already exist, which is closer to how the grid actually gets upgraded.
The same method should carry to solid-state transformers, where converter design, magnetics and control all sit inside one optimization loop.
The general ambition is AI improving the power infrastructure that enables more AI.
4/5
1
20
1,184
General Trajectory retweeted
Introducing Phoenix-DETR: SOTA drone swarm detection on the edge.
Phoenix-DETR is a Detection Transformer using multi-scale deformable attention to detect + track FPV swarms.
The result is strong generalization: up to 63% gains on difficult objects where the best baselines achieve 0%, while maintaining near-perfect performance on standard objects.
General Trajectory retweeted
In just ~3 months, as a solo founder with no prior robotics experience, @gentrajectory trained a foundation model for dexterous manipulation that lets humanoid robots pick up unseen objects and perform real-world work.
It generalizes to novel objects and scenes, including cases where prior SoTA models achieve 0% success.
Congrats on the launch @joshuabelofsky!
ycombinator.com/launches/P6q…
General Trajectory retweeted
General Trajectory (@gentrajectory) is training reasoning models for robotics to bring AI into the physical world. Their first model automates real-world work like mixed-SKU palletization, picking, sorting & packing.
ycombinator.com/launches/MyN…
Congrats on the launch @joshuabelofsky!




