Interesting, how is writing less code cutting costs for them? Does this get back to the rumor that the board was mad at Altman for prioritizing chatgpt over money going into research/model training?