02 MAY 2026
David debugged: Investigated why tool usage analytics showed mock data instead of real counts
David questioned whether the AI tool usage statistics on the redesigned profile were computed from real data or fabricated, discovering the counts were based on incomplete session searches.
During a profile review, David spotted that the tool usage numbers looked wrong and challenged the agent on their source:
what do each of the lines on the claude code graph mean? and why is there nothing on the others? and is this mock data or did you actually compute this? how will it be computed for other users? have you set up the system that re-computes after each write? why does my chatgpt say last seen 1d ago and claude code says last seen 6d ago? again, are you creating a side project mock profile here or are you building an accurate quantitative production product?
The investigation revealed the agent had searched only a subset of session transcripts, missing sessions stored in C:\Users\dkahu\downloads\therapa-upstream on the Windows filesystem. A broader search across both WSL and Windows environments was needed. David's persistent questioning forced an honest accounting of the data's actual coverage.