Is plan mode still worth it?
retained text
Been thinking about this for a while and want to know if others feel the same. When I started using Claude Code seriously, I used plan mode for almost everything. It made sense. The model needed clear instructions or it would go off in a weird direction. Lately I barely use it. The model reads the codebase, figures out the patterns, and makes decisions I would have made anyway. Writing a detailed plan first often takes longer than just letting it try, looking at the result, and correcting it. The other thing: I don't actually read the plans properly. They're long, over-structured, and my eyes skip half of it. I hit approve and hope for the best. That's not planning, that's a ritual. What actually works for me now is a tight loop. Give it the goal, let it do a small chunk, check what it did, push back, repeat. The thinking still happens, it's just spread across the conversation instead of dumped in one doc at the start. But there's a real problem this doesn't solve. When the agent changes a lot of files fast, I lose track of why things are the way they are. Plan mode was a bad fix for that, but at least it was trying. Now I don't have anything. So two questions: 1. Do you still use plan mode? For what kind of tasks? 2. How do you keep a clear picture of your codebase when most of the code is being written by agents?
document fetched 2026-09-29 00:41 · canonical https://reddit.com/r/ClaudeCode/comments/1ws7gvb/is_plan_mode_still_worth_it