Loop Engineering from First Principles — Kyle Mistele, HumanLayer
Production AI agent loops need discipline to avoid unreviable 40,000-line PRs on real teams
“we're building 40,000line PRs that just nobody wants to read”
Kyle Mistele argues that naive prompt-and-loop approaches to agentic coding are producing unmanageable code volumes unsuitable for teams with regulatory and SLA obligations. He advocates for purpose-built loops designed for real-world production constraints rather than solo or greenfield projects. Notably, he cites Claude Code creator Boris Cherny as saying his entire engineering job is now writing loops to prompt Claude, signaling a broader shift in how senior AI engineers conceptualize their work.