Simon Willison finds vibe coding and agentic engineering increasingly blurred as AI coding agents become more reliable

Willison, a software engineer with 25 years of experience, has observed that the distinction he once drew between vibe coding (writing code without understanding or reviewing it) and agentic engineering (using AI tools while applying professional expertise) has started to collapse in his own practice.

As coding agents like Claude Code have become more reliable at producing correct code, Willison has stopped reviewing every line generated for production systems—a shift he initially viewed as irresponsible but now reconciles by treating agents as semi-trusted services, similar to how he trusts internal teams at larger organizations without auditing their code.

The convergence raises a question about accountability: while teams build professional reputations that deter poor work, AI agents cannot take accountability for their output, yet Willison finds them consistently producing correct code in his preferred style, creating what he calls a risk of 'normalization of deviance.'