SkillOrchestra is a new way to make teams of AI models and tools work together by thinking in terms of skills, not just picking one big model for everything.
When you tune the learning rate carefully, plain old LoRA fine-tuning works about as well as fancy new versions.