Hi, thanks for checking out the Startup CTO’s Handbook! This repository has the latest version of the content of the book. You’re welcome and encouraged to contribute issues or pull requests for additions / changes / suggestions / criticisms to be included in future editions. Please feel free to add your name to ACKNOWLEDGEMENTS if you do so.
A novel approach that leverages momentum-based self-distillation to promote consistency and accuracy across the blocks while decoupling the number of blocks from the physical GPU count.
Unlike other models like Sora, Pika2, Veo2, HunyuanVideo’s neural network weights are uncensored and openly distributed, which means they can be run locally under the right circumstances (for example on a consumer 24 GB VRAM GPU) and it can be fine-tuned or used with LoRAs to teach it new concepts.
“Trying to get everyone to license training data is not going to work because that’s not what copyright is about,” Jeffries wrote. “Copyright law is about preventing people from producing exact copies or near exact copies of content and posting it for commercial gain. Period. Anyone who tells you otherwise is lying or simply does not understand how copyright works.”
The AI community is full of people who understand how models work and what they’re capable of, and who are working to improve their systems so that the outputs aren’t full of regurgitated inputs. Google won the Google Books case because it could explain both of these persuasively to judges. But the history of technology law is littered with the remains of companies that were less successful in getting judges to see things their way.
