One calendar year of development on the codebase behind the world’s browsers, measured statement by statement in a single command: 4.53 million logical statements of churn across a 326,948-file tree, its full composition, and how to reproduce every figure.
Chromium is the open-source browser codebase behind Chrome, Edge, Brave, Opera and others — among the largest publicly accessible codebases in existence. We measured one calendar year of its development (snapshots at 31 December 2023 and 31 December 2024) with CodeDelta (codedelta.app) in its git-native mode: one command, both 43.5-million-line snapshots, full CHG/DEL/ADD classification of every logical statement. The year churned 4,534,768 logical statements — 12,390 per day — and touched 143,396 files, 44% of the entire tree. Of all that change, 9.1% was editing of statements that already existed (REWORK); the remainder was insertion and removal. That places Chromium squarely in the mega-project band established by our 1,412-measurement corpus (WebKit, Linux kernel, FreeBSD, IntelliJ: 5–10% REWORK), at 3.2× the Linux kernel’s 2024 churn volume. The scan completed in 105 minutes on consumer hardware — a laptop with 8 GB of memory — with no sampling and no machine learning: deterministic arithmetic, reproducible from the two commit hashes in §05.
We took two year-apart snapshots of the code that runs most of the world’s web browsers and compared every statement. In one year its developers added 2.4 million statements, removed 1.7 million, and edited 411 thousand in place — touching nearly half the files in the project. The striking part is the shape: only about one change in eleven was an edit to existing code. Big, healthy projects at this scale all look like this — our measurements of WebKit, Linux, FreeBSD and IntelliJ show the same pattern — and knowing that pattern is what lets unusual development processes stand out.
All figures are CodeDelta’s own measurements of the 31 December 2024 snapshot (new side of the window):
| Quantity | Measured |
|---|---|
| Total lines (LOC) | 43,491,657 |
| Source lines (SLOC) | 32,171,209 |
| Logical statements (LLOC) | 16,764,119 |
| Files | 326,948 |
| Comment lines | 5,758,295 |
| Source volume | 1.82 GB |
For scale: 24× the largest subject in our nine-codebase study (PostgreSQL, 1.34M SLOC). LOC counts every line; SLOC excludes blanks and comments; LLOC counts logical statements — the unit all churn figures below use.
The measurement is CodeDelta’s standard git-native scan: codedelta --git <old>..<new> extracts both snapshots and computes the full churn classification — CHG (changed in place), DEL (deleted), ADD (added) — per logical statement, per file, per language, using the two-pass sequence-alignment method of Paper A. No sampling, no inference; the arithmetic is deterministic and re-runnable by anyone from public history.
The window is calendar 2024: the last commit of 2023 (502dd460, 31 Dec 2023 23:55 UTC) to the last commit of 2024 (2fc24fd0, 31 Dec 2024 15:57 PST) on main of chromium.googlesource.com/chromium/src. History was fetched shallow from June 2023 (25 GB pack, 7.8 million git objects, ~550,000 commits).
Hardware: a consumer laptop with 8 GB of memory, the repository and all working data on an external USB hard drive. Engine 1.8.6. Wall-clock for the full scan including report generation: about 1 hour 50 minutes — consistent within 3% across four independent runs.
| Churn quantity | LLOC | Share of churn |
|---|---|---|
| Added (ADD) | 2,395,495 | 52.8% |
| Deleted (DEL) | 1,728,106 | 38.1% |
| Edited in place (CHG) | 411,167 | 9.1% |
| Total churn (CRN) | 4,534,768 | 100% |
REWORK = 9.1% (one statement in eleven edited in place); REP_CHURN 0.909. Decomposed by change shape: REWRITE (co-located delete+add pairs) 149,543 · PURE_DEL 1,578,563 · PURE_ADD 2,245,952. Comment churn: 507,853 lines. Physical-line churn: 9,938,827 SLOC.
| File activity | Files |
|---|---|
| Modified | 83,187 |
| Created | 39,312 |
| Deleted | 20,897 |
| Touched (44% of tree) | 143,396 |
| Unchanged | 183,552 |
Rates, for intuition: 12,390 logical statements of churn per calendar day, weekends included; 27,155 physical source lines per day.
Chromium’s REWORK of 9.1% lands inside the band our measurement corpus (1,412 repo-year measurements, 135 repositories, 2010–July 2026) establishes for mega-projects — same-method 2024 windows:
| Codebase | Window | Churn (CRN LLOC) | REWORK |
|---|---|---|---|
| Chromium | 2024 | 4,534,768 | 9.1% |
| WebKit | 2024 | 4,973,778 | 6.0% |
| Linux kernel | 2024 | 1,436,681 | 9.3% |
| FreeBSD | 2024 | 955,664 | 8.2% |
The pattern from Paper C holds at every scale we can find: mid-size established projects rework ~17% of their churn; the very largest projects run 5–10% — more growth surface, more vendored and generated content, more contributors landing new code — and intensively AI-agent-built codebases collapse below 1%. Chromium, at three times the kernel’s churn, sits exactly where the stratum predicts. The measurements are what they are; readers are invited to draw their own conclusions and re-run the window.
codedelta --git 502dd46098ac722e61912aa7b91a3ee370aa0e64..2fc24fd052f0bd7c692ddf375773ef2a58247701 against chromium.googlesource.com/chromium/src. The run record, with full provenance, is retained in the study’s canonical results log.