Standard
A Core Web Vitals metric measuring the delay between a user interaction and the next visual update the browser paints.
INP reports the worst click of a visit. Third-party tags usually own it. Partytown runs those tags in a web worker, where every document.cookie read becomes a blocking message back to the thread you were trying to protect.
Reality
Per-byte web carbon accounting treats hardware manufacturing as a proportion of payload. Embodied carbon is fixed at the factory and spread over however long someone keeps the phone, and page weight has no bearing on how long that is.
Reality
The same component that felt instant on a dozen-entry filter recorded a 1,256 ms local INP on the Shoes brand filter, and Subito fixed it with about 90 lines of virtualization keyed to one measured row height.
Reality
Lighthouse's performance score is a weighted average of five lab metrics whose weights add to 100, and the set leaves out Interaction to Next Paint. The 0 to 100 comes entirely from one simulated load, and field data sits next to it.
Reality
The dev.to writeup behind it argues for keeping every session on infrastructure you own, because a sampled replay tool captures a silent failure only by luck and click-based frustration signals never see scroll or drag.
Reality
A field audit write-up argues red INP scores get answered with hosting migrations, while the main-thread time is actually spent by marketing tags that never went through code review.
Reality