Content studios built around a single model tend to plateau fast — the wins I've seen come from orchestrating 2–3 specialist calls (outline → draft → editor pass with different system prompts) rather than one big mega-prompt. Quality jumps and you get clean checkpoints for human review. What's the underlying stack — are you chaining calls or running a single shot?