NOS (diffusioN Optimized Sampling) is a guidance method for discrete diffusion models that performs antibody CDR sequence design directly in sequence space. It follows gradients in the hidden states of the denoising network to optimise for physicochemical objectives (e.g. SASA, beta-sheet content). Combined with LaMBO-2 Bayesian optimisation it achieves multi-objective antibody optimisation under locality and liability constraints.