ultrawork

Parallel execution coordinates multiple independent agent tasks concurrently. You decompose a request into independent work, select suitable model tiers, run them simultaneously in the background, and verify completed results.

Model tier routing matches simple, standard, and complex work to low, medium, or high-capability agents. This ensures each concurrent task uses an appropriate model tier, optimizing multi-agent orchestration across independent engineering tasks.

Background execution runs long operations such as builds, installations, and test suites without blocking shorter tasks. You can execute long-running operations concurrently while shorter independent tasks proceed simultaneously.

Ask the orchestrator to decompose the request into independent tasks like implementation, test creation, and documentation. Select suitable model tiers for each, execute them concurrently, and apply lightweight verification afterward.

Parallel task delegation requires tasks with no unmet dependencies. You should not use it when tasks depend on each other's output, as simultaneous execution requires independent work to run concurrently without blocking.

Yes, explicit model-tier selection is required for parallel agent execution. You must match simple, standard, and complex work to appropriate low, medium, or high-capability agents before delegating tasks concurrently.