2h ago

Nan Jiang and Modal's Charles Frye call for collaboration on scaling open-source reinforcement learning infrastructure

They target delta compression and weight synchronization challenges

——0——
Original post
Ben (no treats)#982@ANDERSONBCDEFGOPNan JiangNan Jiang@NANJIANGWILL

At @modal, we're working to make sure OSS RL frameworks have all the techniques necessary to train frontier open-weights models. Delta compression is key, but the job's not done. There are still lots of open problems around weight sync, auto-scaling, & cross-cluster training. My DMs are open!

12:56 PM · May 30, 2026 View on X
QUOTE POSTCharles 🎉 Frye#863Charles 🎉 Frye@CHARLES_IRL

why use many bytes when few do trick?

Nan JiangNan Jiang@nanjiangwill

At @modal, we're working to make sure OSS RL frameworks have all the techniques necessary to train frontier open-weights models. Delta compression is key, but the job's not done. There are still lots of open problems around weight sync, auto-scaling, & cross-cluster training. My DMs are open!

7:56 PM · May 30, 2026 · 10.3K Views
7:59 PM · May 30, 2026 · 2.4K Views
REPLYBen (no treats)#982Ben (no treats)@ANDERSONBCDEFG

@charles_irl when me jensen they see

Charles 🎉 FryeCharles 🎉 Frye@charles_irl

why use many bytes when few do trick?

7:59 PM · May 30, 2026 · 2.4K Views
8:20 PM · May 30, 2026 · 445 Views