Leaked Claude Code Setup: Sonnet Builds, Haiku Swarms, Opus Architects
Arindam_1729 · x · 2026-10-09
Arindam shares a leaked Anthropic Claude Code setup whose key insight is assigning each model a different job: Sonnet 5.5 runs the main session and builds, Haiku 5.5 swarms the repo for file discovery/AST scans/grep, Opus 5.5 stays as background architect invoked before big plans, after repeated test failures, and before shipping, and a 16ms router handles mechanical calls. He expands this into a 5-layer setup article for running coding agents in YOLO mode (--dangerously-skip-permissions) faster and safer — saving frontier-model context for moments needing real architectural judgment.
Related event: Leaked Claude Code Config Reveals Multi-Model Role Split(2 posts)→
More from coding & agent
- Wes Bos hacks $50 Amazon photo frames into a family dashboard with a coding agent — natesiggard · 2026-10-09
- Real-time fraud detection in Java: Kafka Streams plus vector similarity, explained — blaizedsouza · 2026-10-09
- Systems programming with AI coding agents: how much code should you still write yourself? — blaizedsouza · 2026-10-09
- Simon Willison ships a blog feature by voice, coding with Codex while cooking dinner — athyuttamre · 2026-10-09
- XState mulls optional state-local context for temp, type-safe data in state machines — DavidKPiano · 2026-10-09
- Unsloth adds fine-tuning to turn Qwen, Llama and Gemma into small local decision models — blaizedsouza · 2026-10-09