An LLM would be more efficient at this task, truth be told.
it could be if a new model was released every day where the training set included everything that happened yesterday.