Reteena

TRIDENT is personalized neural decoding for motor intent—mapping an individual’s neural activity to movement-related signals with models tuned to that person. The goal is reliable, efficient readout for motor interfaces and clinical workflows, not one-size-fits-all decoding.

Lark is a biologically inspired decision-making framework that pairs LLM reasoning with an evolutionary, stakeholder-aware multi-agent system. Plasticity, duplication and maturation, ranked-choice aggregation, and compute-aware optimization produce concise strategies that balance competing priorities without collapsing into a single viewpoint.

Graph-based spatial-transcriptomics models often universally tie predictions over measured spots, which leaves expression undefined between any two nodes. Neural Reaction–Diffusion Operators (NRDO) encode irregular samples into a continuous latent field, evolve it under a learned anisotropic reaction–diffusion equation, and decode expression at arbitrary coordinates. Via Strang splitting, the solver integrates diffusion exactly in the Fourier domain to provide stable, second-order schemes. Across 22 sections from four technologies, NRDO improves masked-region reconstruction over graph, Gaussian-process, neural-field, and imputation baselines. These improvements also survive Holm correction against 4 of 7 baselines, one of which also includes a graph model. Parameter-matched controls attribute the gains to operator structure rather than increased model capacity, and improvements increase with gene-level spatial autocorrelation and vanish for genes that lack spatial structure. We prove that a single stationary snapshot constrains the reaction network to only a Lebesgue-null subset of latent space, which bounds any method of this class that could identify from static tissue. Two limitations remain: comparison against the strongest graph baseline is limited by available specimens, and predicted fields are smoother than the measurements.