Inline edit is a focused AI coding interaction for changing selected code directly, typically within one file. Its value comes from matching narrow tasks with narrow context rather than invoking a broader agent workflow. section 9.2.2
Choose inline edit when the requested change is small and its context is local:
- Use it for a change that can be described in roughly five seconds and touches one file. section 9.2.2
- Prefer tab completion for the next few characters, and switch to agent mode when the task spans files or requires architectural understanding. section 9.2.2
- Keep focused requests focused with mechanisms such as file-specific context rather than broad repository context. section 9.1.3
The five-second rule
Adding one log statement through agent mode may produce the same code as inline edit at roughly two hundred times the cost. A team can make the rule operational by reviewing unnecessary agent sessions and redirecting narrow changes to inline edit. section 9.2.2
Inline edit is not a substitute for broader reasoning. Its efficiency depends on limiting context and output; using a larger mode for a trivial change creates avoidable spend without improving the result. section 9.2.2 section 9.1.3
