parallelization

Parallel task orchestration decomposes work into isolated items with per-worker scoping to minimize touchpoints, then merges outputs safely. Each worker handles a precise piece of work to prevent cross-team conflicts during concurrent development.

Single-message parallelism maximizes concurrency by spawning all worker calls within a single assistant message. This approach allows multiple isolated tasks to execute simultaneously while the orchestrator manages task documents to maintain safety constraints.

Safe git writes require explicit approval before performing repository changes during parallel orchestration. The orchestrator enforces safety constraints and manages task documents to prevent unauthorized or conflicting modifications across worker scopes.

Parallel task management applies directly to multi-service workflows like ETNA Trader where frontend and backend run concurrently. It defines per-worker scope and enforces safety constraints to execute parallel services without increasing merge risk.

Parallel orchestration is not ideal when tasks share heavy touchpoints or require sequential dependencies, as per-work-item scoping assumes isolated work. Concurrency benefits diminish when workers cannot operate independently without cross-team conflicts.

The best way to coordinate developer workers is defining per-worker scope for isolated tasks and spawning all calls in a single message. The orchestrator manages task documents and requires explicit git-writes approval to safely merge outputs.