Shipped build
Kev-olution
A harness for agentic research loops: an agent proposes hypotheses, an evaluator tests them, and the loop keeps whatever scores best. Kev-9B triages each batch before evaluation, and Kev-27B picks a new approach when the score stalls.