/Tech2h ago

Builder Launches Dingbotics To Speed Up RL Training For Robotics

239146.8K

#403

Original post

kache@yacineMTB#403inTech

updated dingbotics site with a model trained in 51 minutes. 2 hours -> 51 minutes

the deepmind baseline i grabbed from mujoco playground is 30 minutes

i am a little compute bound right now. that said, i won't be for long. compute multipliers stack, turn one gpu into four

kache@yacineMTB

I'm launching dingbotics: speeding up RL for robotics

go to dingbotics dot com to check out the latest model I've trained. this will work on your phone's web browser

built with mujoco, nvidia warp, pufferlib, raylib & Rich the Kid

7:01 AM · Jun 12, 2026 · 5K Views

Sentiment

Users are excited about Dingbotics speeding up RL training for robotics because it could enable open-sourcing training loops once policies reach 15 minutes.

Pos

100.0%

Neg

0.0%

1 comments with sentiment.

Cluster Engagement

Posts from X

Most Activity

VIEWS1.8KLIKES12

kache@yacineMTB

I'm going to open source the training loop once I hit 15 minutes for a policy. There's no reason it shouldn't be that fast

kache@yacineMTB

updated dingbotics site with a model trained in 51 minutes. 2 hours -> 51 minutes

the deepmind baseline i grabbed from mujoco playground is 30 minutes

i am a little compute bound right now. that said, i won't be for long. compute multipliers stack, turn one gpu into four

2h1.8K120