The Hard Part of Long-Running Agents Is Not the Model

Описание и применение

Долгосрочные агенты для видеообработки не ограничиваются качеством модели, а зависят от системной архитектуры: хранение состояния, разделение предложений модели и действий кода, контроль бюджета и проверка выполнения.

Задача

Long-running AI agents for video production fail not because of model quality but due to system design flaws that mismanage state, budget, and execution flow.

Результат

By treating the agent as a distributed system with explicit state management, deterministic execution, and robust budgeting, NoodleTomato achieved reliable, resumable, and cost-effective long-running video production.

Рабочий процесс

• 1. Receive brief and create initial plan
• 2. Model proposes shot list, narration, and render steps
• 3. Code validates proposals, checks permissions, reserves budget, writes committed state
• 4. Workers execute tasks, report status to artifact store
• 5. Coordinator remains stateless, wakes on job completion
• 6. After all tasks, run validation and audit outside agent control
• 7. If needed, replay state for debugging without re-running paid jobs

Доказательства

The model did not hallucinate. It reasoned correctly from an incomplete snapshot.

Источник

The model should never grant itself a capability, approve its own output, or author the amount shown on a billing card.

Источник