Every so often, a high-effort latency benchmark appears. Methodology and full-stack coverage varies but they tend to show the same thing:

• Ancient and minimal editors stay <10ms, the "noticeable latency" line below which changes are only noticed through second-order effects.

• Rich browser/Electron-based editors stay <100ms, the "interactive performance drop" line discovered decades ago by IBM and vaguely verified by adtech bounce rates in the last two decades.

Motivation

We're looking for an interaction benchmark and not a typing latency benchmark for a reason.

While looping in language servers (e.g. LSP) isn't a VS Code benchmark per se, autocompletion (e.g. CoPilot) and vector embeddings (e.g. bloop) and similar augments are clearly coming.

Will the next high-effort benchmark target typing and scrolling as before? Will we see a race to stuff the maximum amount of intelligence within the 100ms bound? Or will we finally see local UIs blow past the 100 ms wall for good?

Resolution

The next high-effort (multiple interesting editors benchmarked) reproducible (methodology at least mentions hardware, OS, and measurement points) benchmark in milliseconds will be used as a % resolution.

Whatever measure the benchmark prefers (mean/median/95pct/...) will be used, although centroidal measures will be preferred over maximums and high percentiles. If the number is higher than 100ms the question resolves as YES. If the number is lower than 10ms the question resolves as NO.

If no viable latency benchmarks show up, or VS Code tanks in Google search results for some reason or other, we'll resolve N/A.