loop

FLYWHEEL.md introduces a loop-based framework for agentic coding, where AI agents autonomously ship software but stop at human-gated checkpoints for critical decisions, applying Karpathy's AutoResearch loop to real-world software deployment.

The article compares three open-source AI assistants—Hermes, Loop, and Vellum—focusing on their distinct approaches to memory accumulation and knowledge retention. It highlights Vellum's explicit user approval model as the most reliable for maintaining intentional knowledge states over time.