Steve Yegge criticized developers who merely use Cursor occasionally during his interview with The Pragmatic Engineer. He recommended engineers integrate LLMs more deeply into their workflow, moving beyond simple Q&A interactions. Yegge's AI Coding chart outlines an ultimate goal: Level 8, where developers build custom orchestrators to coordinate multiple AI agents simultaneously.

The Hacker News poster acknowledges this advanced approach wouldn't fly in their current workplace, where they still use "the sorry way" of occasional AI assistance. They're specifically interested in non-greenfield applications—how organizations with existing codebases have implemented agent orchestrators. This reflects a broader industry question about practical AI integration beyond experimental projects.

One comment already pushes back against Yegge's vision, noting they primarily use inline assistants for code modification rather than generation. The comment suggests most developers use AI for refactoring, debugging, and documentation rather than building complex orchestrating systems. This indicates a gap between Yegge's theoretical framework and current practical implementation.

The discussion highlights a tension in AI adoption: between vendor-provided solutions and custom-built orchestration systems. Organizations face challenges integrating advanced AI workflows with existing development lifecycles and team practices. The debate ultimately centers on how deeply AI should be embedded in the software development process versus where human oversight remains essential.