“Legora's CEO believes that focusing on fine-tuning large language models in 2023 was a "waste of time" because general models were improving so rapidly.”