Claude Opus 5.5 early experiments are rolling in. @kevin_t_ngo tested creative writing capabilities with a short story generation task centered on a watermelon. No technical benchmarks yet, but shows the model handling narrative structure and thematic development. Worth watching for more systematic evals comparing creative output quality vs Opus 4 and GPT-4 variants.