Dub's Steven Tey has shared a 15-second launch video for the company's AI Reward Builder that he says was produced in a single attempt with Claude Opus 5.5. The clip recreates Dub's product interface and animates a short walkthrough of a reward campaign.
Tey described the result as another one-shot launch video and praised the model's ability to reproduce the interface and add motion. His post includes the finished clip, but it does not disclose the prompt, preparatory material, generation time or whether any human review occurred before publication. The one-attempt claim is therefore a firsthand account from the product team, not an independent benchmark.
What the video is showing
Dub added AI Reward Builder in August. The feature lets a partner-program operator describe a desired incentive in ordinary language, then generates a reward using Dub's existing schema. Dub's example is a re-engagement campaign aimed at partners who have been enrolled for at least 30 days but produced no leads.
That makes the demo a compact test of a practical design task: matching a real product's visual system closely enough to present a feature, then choreographing the interface into a launch clip. It does not mean Claude built the Reward Builder itself.
Anthropic released Claude Opus 5.5 on September 22, positioning the model for long-running coding and knowledge-work tasks. Tey's video offers a narrower example of the same broader shift: teams are testing frontier models not only for code and analysis, but also for polished product communication. How repeatable the workflow is remains unanswered by a single successful clip.
