sonorama
13 posts

sonorama
@hellosonorama
filmmakers & engineers forging the next gen of creative tools for professionals

Every major generative video platform trying to offer "control" has defaulted to node-based workflows. It’s a massive mistake. Every hour you spend stitching nodes together is an hour stolen from the actual creative process. You shouldn't have to wire a motherboard just to direct a scene.

60 teams are presenting today, and the pressure is high. Over 1,000 are expected to attend live, with more than 10,000 viewing live online.

They say it’s over for Hollywood but movies like Project Hail Mary made 140 million worldwide on opening weekend and Hoppers has made over 242 million worldwide since its release. Where are these AI movies they keep taking about?


I packaged up the "autoresearch" project into a new self-contained minimal repo if people would like to play over the weekend. It's basically nanochat LLM training core stripped down to a single-GPU, one file version of ~630 lines of code, then: - the human iterates on the prompt (.md) - the AI agent iterates on the training code (.py) The goal is to engineer your agents to make the fastest research progress indefinitely and without any of your own involvement. In the image, every dot is a complete LLM training run that lasts exactly 5 minutes. The agent works in an autonomous loop on a git feature branch and accumulates git commits to the training script as it finds better settings (of lower validation loss by the end) of the neural network architecture, the optimizer, all the hyperparameters, etc. You can imagine comparing the research progress of different prompts, different agents, etc. github.com/karpathy/autor… Part code, part sci-fi, and a pinch of psychosis :)






Things that seem important if you’re building software right now: 1. You can write code quickly now 2. Competitors can too 3. So can your users 4. If it takes more than a few clicks an agent should be able to do it 5. Users probably want to use their own agents



