Last year I wrote about my initial experiments with DALL-E 2, OpenAI’s image generation model. I’ve been having an absurd amount of fun playing with its sequel, DALL-E 3 recently. Here are some notes, including a peak under the hood and some notes on the leaked system prompt.