Anonymous usage statistics from the NLP++ Language extension for VS Code. Sections appear here only once there is data behind them, so this page fills out on its own as new releases roll out.
Anonymous usage telemetry from the VisualText VS Code extension — machines, not people.
The two panels share the same days but not the same vertical scale — each is scaled to its own peak (shown in its header), because a download day runs far ahead of an install day. Compare each panel against itself over time, not one panel’s bar heights against the other’s.
11 smaller groups hidden (fewer than 2 machines).
Engine version is blank until the engine has run once in a session.
Latest engine 4.2.53 is too new to appear here yet.
Counts sum each event's occurrences, not rows — batched events would otherwise undercount.
Available from extension 3.12.0 onward.
Starts aren't a denominator. A run fires before the engine begins and can end without a recorded outcome, so started ≠ completed + failed — this is not a failure rate.
A + joins blocks assembled onto Bare Minimum (e.g. Dates and Times+URLs).
(their own analyzer) is a run of an analyzer the user wrote; only analyzers that ship with the extension are named.
Individual events by time — no machine or session identifier, and never file names or content. Use the buttons to filter by type.
How to Read These Numbers
These are machines, not people. One developer working across a laptop, a desktop and a WSL instance counts three times. Nothing on this page identifies an individual machine or session.
Every figure is a lower bound. Telemetry is opt-out, and two independent settings switch it off — VS Code’s own telemetry.telemetryLevel and the extension’s nlp.telemetry.enable. An unknown share of usage is therefore invisible here, and totals should be read as “at least this many”.
Small groups are withheld. Where a breakdown would single out a handful of machines, that row is omitted and the number of omitted rows is stated beneath the table.
Command and language-feature statistics start at version 3.12.0. Earlier releases did not record them, so those tables describe only the machines running 3.12.0 or later and cannot be compared with anything from before it.
Runs started and runs finished are counted separately. A run is recorded when it begins and again when it ends, so the two totals differ whenever work is still in progress or a session ended early. The gap is unfinished work, not failure.
