Receipts
Every number below traces to a real repository. No estimates, no rounding up. Verified September 6, 2026.
Open source
Which repositories appear here is not an editorial choice. The set is exactly the repositories pinned on the author’s GitHub profile at the time of verification, re-read on every refresh — so it changes when the pins change, in both directions.
| Repository | Language | License | Commits | Stars | Created |
|---|---|---|---|---|---|
| scriptree | Python | MIT | 230 | 2 | 2026-04-16 |
| pdfcer | Rust | MIT | 1,476 | 0 | 2026-09-03 |
| pdfcer-gui | Rust | MIT | 760 | 2 | 2026-09-03 |
| combridge | C# | MIT | 19 | 1 | 2026-05-24 |
| swformat | Python | Apache-2.0 | 5 | 0 | 2026-06-12 |
| SWFileFixer | Python | MIT | 4 | 0 | 2026-07-31 |
Test coverage
Hand-measured against local working clones, not GitHub HEAD — see the method note below the table. Verified September 6, 2026.
| Repository | Test files | Test functions | Note |
|---|---|---|---|
| scriptree | 180 | 3,125 | pytest suite; tracked *.py containing a test-function definition |
| pdfcer | 371 | 4,773 | Rust engine + CLI; per-crate tests/ plus inline #[cfg(test)] units |
| pdfcer-gui | 488 | 3,666 | Rust desktop shell; its own README additionally reports 175 driven UI checks in tools/ui-verify, which are NOT counted here |
| swformat | 16 | 175 | pytest harness under test/harness/ |
| SWFileFixer | 13 | 278 | measured for the first time on 2026-09-06; previously reported as null (not measured), never as zero |
| combridge | 0 | 0 | no test suite present — a real zero, not an unmeasured one |
Roughly 12,000 automated tests across the six open-source repositories, as of 2026-09-06. One of the six has no test suite at all — that is stated plainly rather than averaged away. Method: local working clones at D:/Dev/<repo>, git-tracked files only.
Corrections and changes of method
Numbers on this page have been wrong before, and the way they are counted has changed. Both are recorded here rather than quietly overwritten — a page arguing that code should be judged on evidence does not get to edit its own history.
- 2026-08-11 — correction to the figures published 2026-08-10
- The figures published on 2026-08-10 (402 files / 5,783 functions) were WRONG IN BOTH DIRECTIONS. File counts were inflated because the measurement swept each repo's `.claude/` directory, which contains AGENT WORKTREE COPIES of the same codebase. GOTCHA, kept because it generalises: a repository that contains copies of itself will happily let you measure the copies.
- 2026-09-06 — change of measurement method
- The measurement now counts ONLY git-TRACKED files (`git ls-files`), which is a stricter and self-maintaining version of the 2026-08-11 exclusion list: anything ignored — .claude worktrees, target/, node_modules/, .venv/, dist/, build output, scratch files — is excluded by construction rather than by a hand-kept denylist that has to anticipate the next directory nobody thought of. Consequence worth stating plainly: some counts moved without any tests being written or deleted (swformat went 20 files -> 16 because four untracked harness files had been counted; scriptree went 171 -> 180 under a broader file glob). Both directions are reported rather than only the flattering one.
The closed-source platform
ScripTreeApps is not open source — its source is not shown on this site — but the commit history is still honest scale evidence, and the product itself is live and linkable.
Stack: Next.js · Apache OFBiz · Gitea · PostgreSQL · Stripe Connect · Caddy · MinIO · Zoekt