First episode of the show, and the first rung of the ladder: pick a hosted AI video tool, get a clip out of it fast, and learn the one skill that survives every model swap, reading the live leaderboard and benching the top few tools on your own shot. **Tutorial.** We cover the mindset shift from cost-per-generation to cost-per-finished-clip, a light glossary for total beginners (prompt, seed, aspect ratio, text-to-video vs image-to-video, credits), and a copyable first-clip workflow on a free front-end like [Google Flow](https://labs.google/fx/tools/flow): choose text or image to video, set aspect ratio and duration first, write a five-slot prompt (subject, action, environment, light, camera), generate, then score motion, prompt adherence, faces and hands, and value per second, capping yourself at three rolls. Then the durable skill: how to read the [Artificial Analysis Video Arena](https://artificialanalysis.ai/video/arena), why the with-audio and without-audio tabs are different rankings, why the top cluster (not rank 3 vs 4) is your candidate pool, and how to break the tie by benching one representative shot across the top two or three. The pitfall: chasing the monthly leader (the HappyHorse stealth-model saga) and torching credits on blind re-rolls. **News (as of 2026-06-03).** A quiet week for fresh frontier drops. Alibaba's HappyHorse-1.0 sits atop both boards with its "open weights" still half-released; OpenAI is winding Sora down (app gone in April, API ends September), so migrate to Kling, Veo, or Seedance; ByteDance's Seedance 2.0 holds the value crown with 2.5 reportedly coming. Plus a crowded native-audio race (SkyReels V4, PixVerse V6) and the by-job leaderboard snapshot. *AI-generated podcast by OCDevel. Models and prices move monthly, verify the live leaderboard before quoting any ranking.*