Really interested to see what the future of large language models will be, especially since we’ll be able to run them locally on our own hardware and train them on whatever specialized sources we want.