Claude Code dual-model setup: one model plans and reviews, subagents write the code
PrajwalTomar_ · x · 2026-11-21
A practical Claude Code orchestration setup:
- Set /model to Opus 5.5 for execution and /advisor to a planner model that never writes code—only plans, asks questions, and reviews
- Workflow: the advisor grills you one question at a time until the plan is clear, splits it into tickets, and each ticket goes to an Opus 5.5 subagent
- Add three rules to CLAUDE.md: consult the advisor before a plan locks, after the same error twice, and before marking a task done; delete rules written for older models
- The prompt asks for every change as a diff first, with no edits until approved
The author used this to build a full Dynamic Island for his gym app despite not knowing Swift.
More from coding & agent
- Dev argues agent approval flows should let you approve actions by voice — NYCounihan · 2026-11-21
- Agents are sessions, not requests: Google explains sizing GKE node pools for agent workloads — rseroter · 2026-11-21
- Agent Scale Lessons: Shared Specs, Adversarial Pairs, Retiring Long-Run Agents — LukeW · 2026-11-21
- Agent sandbox boat.dev launches at $0.036/hr, hiring founding engineer at $260k — samsja19 · 2026-11-21
- One stray print() inside an MCP stdio server corrupts the JSON-RPC stream — Independent-Sir3234 · 2026-11-21
- Evals for AI Engineers: Shankar and Husain on the Three Gulfs of LLM development — sh_reya · 2026-11-21