Centralized AI drug discovery platforms often reward secrecy over scientific progress—models and data are locked behind NDAs to protect proprietary advantage, teams duplicate one another’s efforts, and exploration can become concentrated around well-trodden chemical scaffolds. This siloed approach breeds redundancy, limits collaboration, and makes it difficult to scale discovery into truly novel regions of the chemical universe. The problem extends beyond access. Drug discovery faces a persistent reproducibility challenge: results that appear promising in one dataset, laboratory, or experimental setting do not always translate to another. Without broader data-sharing, rigorous benchmarking, and independent stress-testing, predictive models can overfit narrow datasets and falter when faced with unfamiliar targets. Even powerful models remain constrained by the data on which they are trained and the candidates they are asked to evaluate. Closed discovery systems can also require substantial capital expenditure to build and maintain proprietary infrastructure, assemble specialized teams, and repeatedly develop capabilities that already exist elsewhere. When each organization must reproduce much of the same infrastructure, scientific progress becomes more expensive than it needs to be. NOVA, our discovery competitions platform, flips the script. By hosting continuous competitions on a decentralized network, we align rewards with transparency and merit while tapping a global network of interdisciplinary talent. Participants compete to identify the highest-scoring molecules and nanobodies across vast chemical and biological spaces, evaluated against state-of-the-art predictive models. Only the best submissions are rewarded, creating persistent pressure to improve the quality and efficiency of discovery. NOVA also hosts competition around the methods used to find candidates. Search algorithms compete to explore chemical space more efficiently, while model competitions push the boundaries of prediction. Winning approaches are open source and can serve as the foundation for others to build on, allowing improvements in one discovery cycle to raise the baseline for the next. This creates a powerful relationship between search and scoring. Better search identifies which regions of chemical space are worth exploring and which candidates merit more expensive evaluation. Better predictive models make that search more informative. As experimental results become available, they can provide the ground truth needed to evaluate and improve both. Over time, NOVA can accumulate more than individual winning submissions: candidate libraries, search algorithms, predictive models, benchmark data, and experimental evidence. These assets can support future discovery programs and make the platform increasingly useful. Our goal is to turn drug discovery from a series of isolated efforts into a continuously improving system—one that can explore more broadly, adapt to new scientific problems, and make the strongest approaches available to the next generation of discovery. NOVA is the discovery engine behind Metanova’s internal pipeline, but its architecture is not limited to our therapeutic thesis. It can be adapted to diverse discovery and R&D objectives, enabling partners to bring their own scientific problems to the platform.