Developer delegates coding to AI agents, keeps product decisions for himself

A product developer building an app called MERIDIAN used AI agents to generate a working prototype in just two days, a task he estimates would have taken a week a year ago. He argues that cheap, fast code generation has shifted the core risk in product development from 'can we build it' to 'are we building the right thing.' In response, he restructured his workflow so agents handle high-volume tasks like market research, competitor analysis, and code writing, while he retains final say on priorities, roadmap decisions, and validating agent outputs. To counter AI overconfidence, he uses a separate agent to critique what the first one builds, and categorizes every claim as known fact, weak signal, or unconfirmed wish. He concludes that product judgment — deciding what is worth building and catching flawed reasoning — is now the irreplaceable human role.