science would suggest to verify deepseek's ability to make ai answers/composition comparison to openai would require they are presented with the same data and form/state/etc and allowed to do their thing witthout using another companies ip and using their reported and inspected hardware/software. then, the llm would be used with various queries and the results evaluated as data scientists think most appropriate, including the contestants. resource utilization would be an important consideration. raw data will have hidden markers/archiving points so the sources are knowable?
Market Data 
Markets 
AI
