Voices
Andrew Ng's Three-Loop Model for AI-Native Product Building: Minutes (Agent), Hours (Developer), Days (Market)
In The Batch, Ng maps AI product development onto three loops running on different clocks — agentic coding in minutes (the agent writes and tests), developer feedback in hours (a human steers), and external feedback in days (users and market decide). The insight is treating these as distinct control loops rather than one blurred workflow, so you know which clock a given decision belongs to. For a solo builder, it's a clean mental model for where to insert human judgment versus where to let the agent run.
↳ Follow the thread