Skip to content

News

Cognition: Kevin-32B: Multi-Turn RL for Writing CUDA Kernels

Product: Cognition

We explore reinforcement learning in a multi-turn setting, using intermediate feedback from the environment, and masking model thoughts to avoid exploding context over multiple tu…

What changed

  • We explore reinforcement learning in a multi-turn setting, using intermediate feedback from the environment, and masking model thoughts to avoid exploding context over multiple tu…

Source

Why it matters

If you use Cognition, skim the official notes above and decide what to try in your next session. Tag the right surface — Cloud vs Desktop vs CLI — before you change workflows.


Fan coverage from Devin Central — not an official Cognition announcement. Devin is a trademark of Cognition.