@benhylak posted that certain video outputs are incredible and linked to a demonstration using the Qwen-3.8 model. The accompanying note states that users refined prompts with lessons from a shared post, allowing the model to produce a full video without extra tools. A related note mentions reinforcement learning techniques applied to the same model. Danielle Fong retweeted the post. The packet records these reactions and generated summaries of the prompting approach but contains no first-party confirmation from the model developers.