empirical-run

To orchestrate an empirical run loop end-to-end, you can automate the worker and redteam cycle through structured phases including preflight validation, question selection, dispatch, evaluation, and verdict handling. The process enforces strict state management and durable updates.

State management during empirical run orchestration works by persisting durable state updates and log trails across cycles in files like state.json, questions.jsonl, truth_ledger.jsonl, and last_redteam.md to ensure reproducible and auditable runs.

Yes, you can pause and resume long-running empirical experiments. The orchestration supports gated pausing and a guarded commit flow to safely manage natural-ceiling pauses and preserve durable state across cycles without losing progress.

The worker redteam loop in empirical run automation is a structured cycle that drives preflight validation, question selection, worker round dispatch, redteam evaluation, and verdict handling to ensure reproducible and auditable experiment trajectories.

You start an empirical run by invoking the command with your specific slug identifier. This initiates the worker and redteam loop, triggering preflight validation before automatically driving the orchestration cycle to completion or a natural ceiling pause.