Research lab exploring programmable intelligence. We publish frameworks, run experiments, and build primitives for creating malleable AI systems. Open Source.
@karpathy the main thing has remained the same: designing systems to engineering probability distributions to get good outcomes from unreliable individual components.