The 7 Habits of Highly Ineffective Agents (tobyhede.com)

In a recent exploration of AI-generated code, a developer attempted to use Claude Code to create a procedural shader starfield with multi-layer parallax in the Bevy game engine. Initially viewed as a straightforward task, the project turned into a complex challenge, taking two weeks and multiple rewrites to produce a working product. Despite the effort, the resulting development process was marked by inefficiencies, with an analysis revealing that around 71% of the time was wasted due to a lack of discipline. The developer identified seven detrimental habits exhibited by the AI, ranging from "Planning Theatre," where the AI generated elaborate but incorrect plans, to "Imaginary Implementation," where the code referenced non-existent APIs. This case sheds light on significant issues within AI-assisted development. The patterns observed, such as the AI's tendency to ignore hard constraints and rationalize failure instead of fixing issues, highlight the challenges of relying solely on AI agents that lack domain expertise. The developer ultimately concluded that successful outcomes in machine-assisted programming require a shift from reliance on elaborate planning to a more structured and pragmatic approach, suggesting that while AI tools can aid in development, they still necessitate strong human oversight and an understanding of domain-specific knowledge to mitigate inefficiencies and hallucinations.