Senior engineer, loop orchestrator sample setup

A senior engineer shared a sample setup for a loop orchestrator, detailing how to choose AI models for coding tasks based on complexity. For COMPLEX tasks, such as those involving multi-file, concurrency, data, persistence, security, architecture, or high-stakes correctness, the 'claude-fable-5' model should be used. LESS COMPLEX tasks, including small, cosmetic, mechanical, or documentation work, should utilize the 'claude-opus-4-8' model. The model must be explicitly passed on 'create_session', and reviewers and spikes follow the same complexity-based rule.

This setup uniquely specifies different Claude models for tasks based on complexity, unlike general recommendations that often suggest a single model.