We used parametrically designed proteins that are small, easily expressed, and very stable. We used their sequence in a custom boltz2-solubleMPNN iterative redesign cycle. We only accepted boltz2 prediction that increased in ipSAE_min for the next generation of designs. We found that changing 10% of residues to X in random positions kickstarts this process of refinement. Ultimately the fold and core of these miniproteins remains mostly unchanged while ipSAE_min could go up to 0.8. We expect high success rate in expressibility and stability of these binders. We borrowed code from ProteinHunter to run Boltz and sollubleMPNN.