Choose AI Coding Models by Delegation Level, Not Raw Intelligence

A commentary piece argues that highly autonomous AI coding tools like Fable and Claude Opus may not suit engineers who have strong, established development preferences. The author draws on Simon Willison's observations to suggest that experienced developers often prefer steering a mid-tier model like Sonnet closely, rather than delegating broad decisions to a flagship model. The piece uses a manual-versus-automatic car analogy to illustrate how personal coding style influences which type of AI tool feels most natural. For developers without a fixed workflow, or those tackling large implementation tasks, delegating broadly to an autonomous model can be more efficient. The key takeaway is that model selection should be based on how much control a developer wants to retain over the development process, not simply on which model is most capable.