Antoine van der Lee • 3/23/2026

The “One More Prompt” risk of agentic coding

This article discusses the 'One More Prompt' risk in agentic coding, where developers using AI agents are tempted to continuously add features or fixes via additional prompts, resulting in overly large pull requests. The author shares a personal experience building a post-editor for RocketSim's Xcode Simulator recording feature, highlighting how easy prompting led to a PR with 13,758 lines added. The article warns against scope creep and emphasizes disciplined review and scoping to avoid this trap. It is relevant to IT/technology, focusing on software development practices with AI agents.

0 comments

#ai-assisted development #code generation #Development Workflow