work on creating gpt-zero from ground-up Learned about how to code QKV attention mechanism, batching, and pretraining