Accelerated Understanding Inc is launching today. Their model uses a non-transformer architecture called neural operators, and this appears to have been the source of recent rumors and about a new model with an extremely large context window. Release announcement incoming.