Designing molecules that bind was never the hard part

What we believe about biomolecular design, and why LiteFold is built the way it is. Six positions, held until the evidence says otherwise.

1. Binding is a starting condition, not a result. What kills molecules downstream is selectivity, stability in serum, solubility at dose and a synthesis route that exists.
2. Objectives should be optimized jointly, not in sequence. We treat the objectives as one surface and search it directly, so trade-offs are made where they are visible.
3. A model without an environment cannot improve. We build oracles, simulators and structural checks a sampler can query mid-generation.
4. Modality is an implementation detail. Our models generate across small molecules, peptides, macrocycles, PROTACs and non-canonical residues.
5. The wet lab is part of the model. We work with partners who synthesize and measure, and route those measurements back into the loop.
6. Publish the thing that can be checked. We release pre-prints, benchmarks, datasets and the failure modes that came with them.