Kief Morris, writing on Martin Fowler's blog, argues developers are unlikely to go fully 'out of the loop' in AI-assisted development. Instead, a third model — working 'on the loop' — is emerging, where developers design tests, specifications, and feedback mechanisms that guide AI agents rather than reviewing every generated artifact. Industry examples from OpenAI's Codex agent loop and Datadog's 'harness-first' approach illustrate this shift. Stack Overflow's 2025 survey shows 84% of developers use or plan to use AI tools, but trust in AI-generated output remains low, with concerns around technical debt and maintainability. The consensus is that automated verification pipelines, evaluation frameworks, and observability systems are becoming essential as coding agents grow more capable.
Sort: