📖 Read the full article: Testing Fable vs Sol in terms of taste
Per-run artifacts from the 8 video-production runs behind that article — for each task × model, the script, director's statement, and (for the grounded tasks) the researched fact sheet, exactly as each model wrote them.
videos/<task>-<model>.mp4 the 8 final cuts
posters/<task>-<model>.jpg poster frames
files/<task>-<model>/
script.md the shooting script the model wrote and shot
director_statement.md the model's own explanation of why it stopped
facts.md (bigben, codex-micro) web-researched facts the run grounded on
research.md (some runs) the model's product/market research
The full harness (briefs, skills, runner, summary tooling) is a separate repo — see the article.