Hey, thanks for the detailed notes. I’ll reply point by point:
Lines means lines of code added plus removed in AI suggestions.
Numerator and denominator: the denominator is lines the AI suggested for the current day Today, the numerator is lines you accepted or kept. The percent is the acceptance rate.
Tab AI Stats counts autocomplete suggestions from Tab.
Agent AI Stats counts diffs from Agent or Composer. The numbers can differ between modes, likely because some edits get applied differently, for example some auto applied changes might not be counted as accepted lines. That’s my guess though since the docs don’t explain the exact mechanics.
These are informational productivity metrics. They are not related to usage limits, quota, or billing. For limits and spend, use the separate Usage Summary. They’re different on purpose so it’s easier not to mix them up.
There isn’t separate documentation for these status bar metrics right now.
Why the percent can be over 100% like 217%: the two counters update from different events and don’t match 1:1, so accepted lines can end up higher than suggested lines. This is a known anomaly and your observation is correct. Similar reports exist for API acceptance rate: https://forum.cursor.com/t/more-accepted-than-suggested-lines/130872
If these elements get in the way, you can usually hide them by right clicking the status bar. That’s standard status bar behavior, not guaranteed for these new items, but it’s worth trying.
That clears up my confusion. I initially thought it might be related to usage limits because the percentage changed so dramatically and even exceeded 100%. Good to know it’s just a productivity metric. I appreciate the explanation and the link!