I’ve always wanted to do an analysis of how I use Claude Code. Then I found out that someone had already done it. I looked at my own usage data, and for now, there’s nothing particularly interesting. I wonder if anyone has any ideas.
For example, error type distribution or intervention rates. This can tell me how efficient I am when using Claude.
But currently, the error type is a bit too broad, and I haven’t discovered much yet.