The CSS that styles marketing pages now ships inline with the HTML instead of as a separate file the browser has to fetch before rendering anything. On a slow connection, that's about a second of load time gone — most noticeable on the demo page where paid traffic lands. The total CSS footprint also dropped by 15KB after notification-related styles were stopped from loading on pages that never use them.
Visibility scans now default to newer, lower-cost models across ChatGPT and Claude. The results are equivalent — you can still override to the previous models if you want — but this keeps scan costs down as usage scales.
At peak, the backend was spinning up roughly one processing job per log drain delivery. Those now batch together, cutting internal job counts by around 100× without any change to how quickly data appears in your dashboard.