In this case though the compute per problem was pretty reasonable so if this model was available this rate of progress would mostly continue whether OpenAI funds it or not.